blob: 50ec42f170a06713e086d66abb2ea0d1aa998d74 [file] [log] [blame]
Jesse Grossccb13522011-10-25 19:26:31 -07001/*
Andy Zhou971427f32014-09-15 19:37:25 -07002 * Copyright (c) 2007-2014 Nicira, Inc.
Jesse Grossccb13522011-10-25 19:26:31 -07003 *
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.
7 *
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.
12 *
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
16 * 02110-1301, USA
17 */
18
Jesse Grossccb13522011-10-25 19:26:31 -070019#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>
30#include <linux/in.h>
31#include <linux/rcupdate.h>
32#include <linux/if_arp.h>
Jesse Grossccb13522011-10-25 19:26:31 -070033#include <linux/ip.h>
34#include <linux/ipv6.h>
Simon Horman25cd9ba2014-10-06 05:05:13 -070035#include <linux/mpls.h>
Joe Stringera175a722013-08-22 12:30:48 -070036#include <linux/sctp.h>
Pravin B Shelare298e502013-10-29 17:22:21 -070037#include <linux/smp.h>
Jesse Grossccb13522011-10-25 19:26:31 -070038#include <linux/tcp.h>
39#include <linux/udp.h>
40#include <linux/icmp.h>
41#include <linux/icmpv6.h>
42#include <linux/rculist.h>
43#include <net/ip.h>
Pravin B Shelar7d5437c2013-06-17 17:50:18 -070044#include <net/ip_tunnels.h>
Jesse Grossccb13522011-10-25 19:26:31 -070045#include <net/ipv6.h>
Simon Horman25cd9ba2014-10-06 05:05:13 -070046#include <net/mpls.h>
Jesse Grossccb13522011-10-25 19:26:31 -070047#include <net/ndisc.h>
48
Pravin B Shelar83c8df22014-09-15 19:20:31 -070049#include "datapath.h"
50#include "flow.h"
51#include "flow_netlink.h"
52
Pravin B Shelare6445712013-10-03 18:16:47 -070053u64 ovs_flow_used_time(unsigned long flow_jiffies)
Andy Zhou03f0d912013-08-07 20:01:00 -070054{
Pravin B Shelare6445712013-10-03 18:16:47 -070055 struct timespec cur_ts;
56 u64 cur_ms, idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070057
Pravin B Shelare6445712013-10-03 18:16:47 -070058 ktime_get_ts(&cur_ts);
59 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
60 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
61 cur_ts.tv_nsec / NSEC_PER_MSEC;
Andy Zhou03f0d912013-08-07 20:01:00 -070062
Pravin B Shelare6445712013-10-03 18:16:47 -070063 return cur_ms - idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070064}
65
Jarno Rajahalmedf23e9f2013-10-23 01:40:44 -070066#define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
Andy Zhou03f0d912013-08-07 20:01:00 -070067
Ben Pfaffad552002014-05-06 16:48:38 -070068void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags,
Thomas Graf12eb18f2014-11-06 06:58:52 -080069 const struct sk_buff *skb)
Andy Zhou5828cd92013-08-27 13:02:21 -070070{
Pravin B Shelare298e502013-10-29 17:22:21 -070071 struct flow_stats *stats;
Jarno Rajahalme63e79592014-03-27 12:42:54 -070072 int node = numa_node_id();
Jiri Pirkodf8a39d2015-01-13 17:13:44 +010073 int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Andy Zhou5828cd92013-08-27 13:02:21 -070074
Jarno Rajahalme63e79592014-03-27 12:42:54 -070075 stats = rcu_dereference(flow->stats[node]);
Pravin B Shelare298e502013-10-29 17:22:21 -070076
Jarno Rajahalme63e79592014-03-27 12:42:54 -070077 /* Check if already have node-specific stats. */
78 if (likely(stats)) {
79 spin_lock(&stats->lock);
80 /* Mark if we write on the pre-allocated stats. */
81 if (node == 0 && unlikely(flow->stats_last_writer != node))
82 flow->stats_last_writer = node;
83 } else {
84 stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */
85 spin_lock(&stats->lock);
86
87 /* If the current NUMA-node is the only writer on the
88 * pre-allocated stats keep using them.
89 */
90 if (unlikely(flow->stats_last_writer != node)) {
91 /* A previous locker may have already allocated the
92 * stats, so we need to check again. If node-specific
93 * stats were already allocated, we update the pre-
94 * allocated stats as we have already locked them.
95 */
96 if (likely(flow->stats_last_writer != NUMA_NO_NODE)
Andreea-Cristina Bernat8c6b00c2014-08-17 16:29:43 +030097 && likely(!rcu_access_pointer(flow->stats[node]))) {
Jarno Rajahalme63e79592014-03-27 12:42:54 -070098 /* Try to allocate node-specific stats. */
99 struct flow_stats *new_stats;
100
101 new_stats =
102 kmem_cache_alloc_node(flow_stats_cache,
103 GFP_THISNODE |
104 __GFP_NOMEMALLOC,
105 node);
106 if (likely(new_stats)) {
107 new_stats->used = jiffies;
108 new_stats->packet_count = 1;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800109 new_stats->byte_count = len;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700110 new_stats->tcp_flags = tcp_flags;
111 spin_lock_init(&new_stats->lock);
112
113 rcu_assign_pointer(flow->stats[node],
114 new_stats);
115 goto unlock;
116 }
117 }
118 flow->stats_last_writer = node;
119 }
120 }
121
Pravin B Shelare298e502013-10-29 17:22:21 -0700122 stats->used = jiffies;
123 stats->packet_count++;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800124 stats->byte_count += len;
Pravin B Shelare298e502013-10-29 17:22:21 -0700125 stats->tcp_flags |= tcp_flags;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700126unlock:
Pravin B Shelare298e502013-10-29 17:22:21 -0700127 spin_unlock(&stats->lock);
128}
129
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700130/* Must be called with rcu_read_lock or ovs_mutex. */
131void ovs_flow_stats_get(const struct sw_flow *flow,
132 struct ovs_flow_stats *ovs_stats,
Pravin B Shelare298e502013-10-29 17:22:21 -0700133 unsigned long *used, __be16 *tcp_flags)
134{
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700135 int node;
Pravin B Shelare298e502013-10-29 17:22:21 -0700136
137 *used = 0;
138 *tcp_flags = 0;
139 memset(ovs_stats, 0, sizeof(*ovs_stats));
140
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700141 for_each_node(node) {
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700142 struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[node]);
Pravin B Shelare298e502013-10-29 17:22:21 -0700143
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700144 if (stats) {
145 /* Local CPU may write on non-local stats, so we must
146 * block bottom-halves here.
147 */
148 spin_lock_bh(&stats->lock);
149 if (!*used || time_after(stats->used, *used))
150 *used = stats->used;
151 *tcp_flags |= stats->tcp_flags;
152 ovs_stats->n_packets += stats->packet_count;
153 ovs_stats->n_bytes += stats->byte_count;
154 spin_unlock_bh(&stats->lock);
155 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700156 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700157}
158
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700159/* Called with ovs_mutex. */
Pravin B Shelare298e502013-10-29 17:22:21 -0700160void ovs_flow_stats_clear(struct sw_flow *flow)
161{
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700162 int node;
Pravin B Shelare298e502013-10-29 17:22:21 -0700163
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700164 for_each_node(node) {
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700165 struct flow_stats *stats = ovsl_dereference(flow->stats[node]);
Jarno Rajahalme23dabf82014-03-27 12:35:23 -0700166
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700167 if (stats) {
168 spin_lock_bh(&stats->lock);
169 stats->used = 0;
170 stats->packet_count = 0;
171 stats->byte_count = 0;
172 stats->tcp_flags = 0;
173 spin_unlock_bh(&stats->lock);
174 }
175 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700176}
177
Jesse Grossccb13522011-10-25 19:26:31 -0700178static int check_header(struct sk_buff *skb, int len)
179{
180 if (unlikely(skb->len < len))
181 return -EINVAL;
182 if (unlikely(!pskb_may_pull(skb, len)))
183 return -ENOMEM;
184 return 0;
185}
186
187static bool arphdr_ok(struct sk_buff *skb)
188{
189 return pskb_may_pull(skb, skb_network_offset(skb) +
190 sizeof(struct arp_eth_header));
191}
192
193static int check_iphdr(struct sk_buff *skb)
194{
195 unsigned int nh_ofs = skb_network_offset(skb);
196 unsigned int ip_len;
197 int err;
198
199 err = check_header(skb, nh_ofs + sizeof(struct iphdr));
200 if (unlikely(err))
201 return err;
202
203 ip_len = ip_hdrlen(skb);
204 if (unlikely(ip_len < sizeof(struct iphdr) ||
205 skb->len < nh_ofs + ip_len))
206 return -EINVAL;
207
208 skb_set_transport_header(skb, nh_ofs + ip_len);
209 return 0;
210}
211
212static bool tcphdr_ok(struct sk_buff *skb)
213{
214 int th_ofs = skb_transport_offset(skb);
215 int tcp_len;
216
217 if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
218 return false;
219
220 tcp_len = tcp_hdrlen(skb);
221 if (unlikely(tcp_len < sizeof(struct tcphdr) ||
222 skb->len < th_ofs + tcp_len))
223 return false;
224
225 return true;
226}
227
228static bool udphdr_ok(struct sk_buff *skb)
229{
230 return pskb_may_pull(skb, skb_transport_offset(skb) +
231 sizeof(struct udphdr));
232}
233
Joe Stringera175a722013-08-22 12:30:48 -0700234static bool sctphdr_ok(struct sk_buff *skb)
235{
236 return pskb_may_pull(skb, skb_transport_offset(skb) +
237 sizeof(struct sctphdr));
238}
239
Jesse Grossccb13522011-10-25 19:26:31 -0700240static bool icmphdr_ok(struct sk_buff *skb)
241{
242 return pskb_may_pull(skb, skb_transport_offset(skb) +
243 sizeof(struct icmphdr));
244}
245
Andy Zhou03f0d912013-08-07 20:01:00 -0700246static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700247{
248 unsigned int nh_ofs = skb_network_offset(skb);
249 unsigned int nh_len;
250 int payload_ofs;
251 struct ipv6hdr *nh;
252 uint8_t nexthdr;
253 __be16 frag_off;
254 int err;
255
Jesse Grossccb13522011-10-25 19:26:31 -0700256 err = check_header(skb, nh_ofs + sizeof(*nh));
257 if (unlikely(err))
258 return err;
259
260 nh = ipv6_hdr(skb);
261 nexthdr = nh->nexthdr;
262 payload_ofs = (u8 *)(nh + 1) - skb->data;
263
264 key->ip.proto = NEXTHDR_NONE;
265 key->ip.tos = ipv6_get_dsfield(nh);
266 key->ip.ttl = nh->hop_limit;
267 key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
268 key->ipv6.addr.src = nh->saddr;
269 key->ipv6.addr.dst = nh->daddr;
270
271 payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off);
272 if (unlikely(payload_ofs < 0))
273 return -EINVAL;
274
275 if (frag_off) {
276 if (frag_off & htons(~0x7))
277 key->ip.frag = OVS_FRAG_TYPE_LATER;
278 else
279 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700280 } else {
281 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700282 }
283
284 nh_len = payload_ofs - nh_ofs;
285 skb_set_transport_header(skb, nh_ofs + nh_len);
286 key->ip.proto = nexthdr;
287 return nh_len;
288}
289
290static bool icmp6hdr_ok(struct sk_buff *skb)
291{
292 return pskb_may_pull(skb, skb_transport_offset(skb) +
293 sizeof(struct icmp6hdr));
294}
295
Jesse Grossccb13522011-10-25 19:26:31 -0700296static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
297{
298 struct qtag_prefix {
299 __be16 eth_type; /* ETH_P_8021Q */
300 __be16 tci;
301 };
302 struct qtag_prefix *qp;
303
304 if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)))
305 return 0;
306
307 if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) +
308 sizeof(__be16))))
309 return -ENOMEM;
310
311 qp = (struct qtag_prefix *) skb->data;
312 key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT);
313 __skb_pull(skb, sizeof(struct qtag_prefix));
314
315 return 0;
316}
317
318static __be16 parse_ethertype(struct sk_buff *skb)
319{
320 struct llc_snap_hdr {
321 u8 dsap; /* Always 0xAA */
322 u8 ssap; /* Always 0xAA */
323 u8 ctrl;
324 u8 oui[3];
325 __be16 ethertype;
326 };
327 struct llc_snap_hdr *llc;
328 __be16 proto;
329
330 proto = *(__be16 *) skb->data;
331 __skb_pull(skb, sizeof(__be16));
332
Simon Hormane5c5d222013-03-28 13:38:25 +0900333 if (ntohs(proto) >= ETH_P_802_3_MIN)
Jesse Grossccb13522011-10-25 19:26:31 -0700334 return proto;
335
336 if (skb->len < sizeof(struct llc_snap_hdr))
337 return htons(ETH_P_802_2);
338
339 if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
340 return htons(0);
341
342 llc = (struct llc_snap_hdr *) skb->data;
343 if (llc->dsap != LLC_SAP_SNAP ||
344 llc->ssap != LLC_SAP_SNAP ||
345 (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
346 return htons(ETH_P_802_2);
347
348 __skb_pull(skb, sizeof(struct llc_snap_hdr));
Rich Lane17b682a2013-02-19 11:10:30 -0800349
Simon Hormane5c5d222013-03-28 13:38:25 +0900350 if (ntohs(llc->ethertype) >= ETH_P_802_3_MIN)
Rich Lane17b682a2013-02-19 11:10:30 -0800351 return llc->ethertype;
352
353 return htons(ETH_P_802_2);
Jesse Grossccb13522011-10-25 19:26:31 -0700354}
355
356static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
Andy Zhou03f0d912013-08-07 20:01:00 -0700357 int nh_len)
Jesse Grossccb13522011-10-25 19:26:31 -0700358{
359 struct icmp6hdr *icmp = icmp6_hdr(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700360
361 /* The ICMPv6 type and code fields use the 16-bit transport port
362 * fields, so we need to store them in 16-bit network byte order.
363 */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700364 key->tp.src = htons(icmp->icmp6_type);
365 key->tp.dst = htons(icmp->icmp6_code);
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700366 memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd));
Jesse Grossccb13522011-10-25 19:26:31 -0700367
368 if (icmp->icmp6_code == 0 &&
369 (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
370 icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
371 int icmp_len = skb->len - skb_transport_offset(skb);
372 struct nd_msg *nd;
373 int offset;
374
Jesse Grossccb13522011-10-25 19:26:31 -0700375 /* In order to process neighbor discovery options, we need the
376 * entire packet.
377 */
378 if (unlikely(icmp_len < sizeof(*nd)))
Andy Zhou03f0d912013-08-07 20:01:00 -0700379 return 0;
380
381 if (unlikely(skb_linearize(skb)))
382 return -ENOMEM;
Jesse Grossccb13522011-10-25 19:26:31 -0700383
384 nd = (struct nd_msg *)skb_transport_header(skb);
385 key->ipv6.nd.target = nd->target;
Jesse Grossccb13522011-10-25 19:26:31 -0700386
387 icmp_len -= sizeof(*nd);
388 offset = 0;
389 while (icmp_len >= 8) {
390 struct nd_opt_hdr *nd_opt =
391 (struct nd_opt_hdr *)(nd->opt + offset);
392 int opt_len = nd_opt->nd_opt_len * 8;
393
394 if (unlikely(!opt_len || opt_len > icmp_len))
Andy Zhou03f0d912013-08-07 20:01:00 -0700395 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700396
397 /* Store the link layer address if the appropriate
398 * option is provided. It is considered an error if
399 * the same link layer option is specified twice.
400 */
401 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
402 && opt_len == 8) {
403 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
404 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800405 ether_addr_copy(key->ipv6.nd.sll,
406 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700407 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
408 && opt_len == 8) {
409 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
410 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800411 ether_addr_copy(key->ipv6.nd.tll,
412 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700413 }
414
415 icmp_len -= opt_len;
416 offset += opt_len;
417 }
418 }
419
Andy Zhou03f0d912013-08-07 20:01:00 -0700420 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700421
422invalid:
423 memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
424 memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
425 memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
426
Andy Zhou03f0d912013-08-07 20:01:00 -0700427 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700428}
429
430/**
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700431 * key_extract - extracts a flow key from an Ethernet frame.
Jesse Grossccb13522011-10-25 19:26:31 -0700432 * @skb: sk_buff that contains the frame, with skb->data pointing to the
433 * Ethernet header
Jesse Grossccb13522011-10-25 19:26:31 -0700434 * @key: output flow key
Jesse Grossccb13522011-10-25 19:26:31 -0700435 *
436 * The caller must ensure that skb->len >= ETH_HLEN.
437 *
438 * Returns 0 if successful, otherwise a negative errno value.
439 *
440 * Initializes @skb header pointers as follows:
441 *
442 * - skb->mac_header: the Ethernet header.
443 *
444 * - skb->network_header: just past the Ethernet header, or just past the
445 * VLAN header, to the first byte of the Ethernet payload.
446 *
Lorand Jakab34d94f22013-06-03 10:01:14 -0700447 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
Jesse Grossccb13522011-10-25 19:26:31 -0700448 * on output, then just past the IP header, if one is present and
449 * of a correct length, otherwise the same as skb->network_header.
Lorand Jakab34d94f22013-06-03 10:01:14 -0700450 * For other key->eth.type values it is left untouched.
Jesse Grossccb13522011-10-25 19:26:31 -0700451 */
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700452static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700453{
Andy Zhou03f0d912013-08-07 20:01:00 -0700454 int error;
Jesse Grossccb13522011-10-25 19:26:31 -0700455 struct ethhdr *eth;
456
Jesse Grossf5796682014-10-03 15:35:33 -0700457 /* Flags are always used as part of stats */
458 key->tp.flags = 0;
459
Jesse Grossccb13522011-10-25 19:26:31 -0700460 skb_reset_mac_header(skb);
461
462 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
463 * header in the linear data area.
464 */
465 eth = eth_hdr(skb);
Joe Perches8c63ff02014-02-18 11:15:45 -0800466 ether_addr_copy(key->eth.src, eth->h_source);
467 ether_addr_copy(key->eth.dst, eth->h_dest);
Jesse Grossccb13522011-10-25 19:26:31 -0700468
469 __skb_pull(skb, 2 * ETH_ALEN);
Pravin B Shelarb34df5e2013-06-13 11:11:44 -0700470 /* We are going to push all headers that we pull, so no need to
471 * update skb->csum here.
472 */
Jesse Grossccb13522011-10-25 19:26:31 -0700473
Jesse Gross07148122014-10-03 15:35:29 -0700474 key->eth.tci = 0;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +0100475 if (skb_vlan_tag_present(skb))
Jesse Grossccb13522011-10-25 19:26:31 -0700476 key->eth.tci = htons(skb->vlan_tci);
477 else if (eth->h_proto == htons(ETH_P_8021Q))
478 if (unlikely(parse_vlan(skb, key)))
479 return -ENOMEM;
480
481 key->eth.type = parse_ethertype(skb);
482 if (unlikely(key->eth.type == htons(0)))
483 return -ENOMEM;
484
485 skb_reset_network_header(skb);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700486 skb_reset_mac_len(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700487 __skb_push(skb, skb->data - skb_mac_header(skb));
488
489 /* Network layer. */
490 if (key->eth.type == htons(ETH_P_IP)) {
491 struct iphdr *nh;
492 __be16 offset;
493
Jesse Grossccb13522011-10-25 19:26:31 -0700494 error = check_iphdr(skb);
495 if (unlikely(error)) {
Jesse Gross07148122014-10-03 15:35:29 -0700496 memset(&key->ip, 0, sizeof(key->ip));
497 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700498 if (error == -EINVAL) {
499 skb->transport_header = skb->network_header;
500 error = 0;
501 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700502 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700503 }
504
505 nh = ip_hdr(skb);
506 key->ipv4.addr.src = nh->saddr;
507 key->ipv4.addr.dst = nh->daddr;
508
509 key->ip.proto = nh->protocol;
510 key->ip.tos = nh->tos;
511 key->ip.ttl = nh->ttl;
512
513 offset = nh->frag_off & htons(IP_OFFSET);
514 if (offset) {
515 key->ip.frag = OVS_FRAG_TYPE_LATER;
Andy Zhou03f0d912013-08-07 20:01:00 -0700516 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700517 }
518 if (nh->frag_off & htons(IP_MF) ||
Jesse Gross07148122014-10-03 15:35:29 -0700519 skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
Jesse Grossccb13522011-10-25 19:26:31 -0700520 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Jesse Gross07148122014-10-03 15:35:29 -0700521 else
522 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700523
524 /* Transport layer. */
525 if (key->ip.proto == IPPROTO_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700526 if (tcphdr_ok(skb)) {
527 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700528 key->tp.src = tcp->source;
529 key->tp.dst = tcp->dest;
530 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700531 } else {
532 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700533 }
Jesse Gross07148122014-10-03 15:35:29 -0700534
Jesse Grossccb13522011-10-25 19:26:31 -0700535 } else if (key->ip.proto == IPPROTO_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700536 if (udphdr_ok(skb)) {
537 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700538 key->tp.src = udp->source;
539 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700540 } else {
541 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700542 }
Joe Stringera175a722013-08-22 12:30:48 -0700543 } else if (key->ip.proto == IPPROTO_SCTP) {
544 if (sctphdr_ok(skb)) {
545 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700546 key->tp.src = sctp->source;
547 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700548 } else {
549 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700550 }
Jesse Grossccb13522011-10-25 19:26:31 -0700551 } else if (key->ip.proto == IPPROTO_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700552 if (icmphdr_ok(skb)) {
553 struct icmphdr *icmp = icmp_hdr(skb);
554 /* The ICMP type and code fields use the 16-bit
555 * transport port fields, so we need to store
556 * them in 16-bit network byte order. */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700557 key->tp.src = htons(icmp->type);
558 key->tp.dst = htons(icmp->code);
Jesse Gross07148122014-10-03 15:35:29 -0700559 } else {
560 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700561 }
562 }
563
Jesse Gross07148122014-10-03 15:35:29 -0700564 } else if (key->eth.type == htons(ETH_P_ARP) ||
565 key->eth.type == htons(ETH_P_RARP)) {
Jesse Grossccb13522011-10-25 19:26:31 -0700566 struct arp_eth_header *arp;
Li RongQing389f4892014-10-17 14:03:08 +0800567 bool arp_available = arphdr_ok(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700568
569 arp = (struct arp_eth_header *)skb_network_header(skb);
570
Li RongQing389f4892014-10-17 14:03:08 +0800571 if (arp_available &&
Jesse Gross07148122014-10-03 15:35:29 -0700572 arp->ar_hrd == htons(ARPHRD_ETHER) &&
573 arp->ar_pro == htons(ETH_P_IP) &&
574 arp->ar_hln == ETH_ALEN &&
575 arp->ar_pln == 4) {
Jesse Grossccb13522011-10-25 19:26:31 -0700576
577 /* We only match on the lower 8 bits of the opcode. */
578 if (ntohs(arp->ar_op) <= 0xff)
579 key->ip.proto = ntohs(arp->ar_op);
Jesse Gross07148122014-10-03 15:35:29 -0700580 else
581 key->ip.proto = 0;
582
Mehak Mahajand04d3822012-10-30 15:50:28 -0700583 memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
584 memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
Joe Perches8c63ff02014-02-18 11:15:45 -0800585 ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
586 ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
Jesse Gross07148122014-10-03 15:35:29 -0700587 } else {
588 memset(&key->ip, 0, sizeof(key->ip));
589 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700590 }
Simon Horman25cd9ba2014-10-06 05:05:13 -0700591 } else if (eth_p_mpls(key->eth.type)) {
592 size_t stack_len = MPLS_HLEN;
593
594 /* In the presence of an MPLS label stack the end of the L2
595 * header and the beginning of the L3 header differ.
596 *
597 * Advance network_header to the beginning of the L3
598 * header. mac_len corresponds to the end of the L2 header.
599 */
600 while (1) {
601 __be32 lse;
602
603 error = check_header(skb, skb->mac_len + stack_len);
604 if (unlikely(error))
605 return 0;
606
607 memcpy(&lse, skb_network_header(skb), MPLS_HLEN);
608
609 if (stack_len == MPLS_HLEN)
610 memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN);
611
612 skb_set_network_header(skb, skb->mac_len + stack_len);
613 if (lse & htonl(MPLS_LS_S_MASK))
614 break;
615
616 stack_len += MPLS_HLEN;
617 }
Jesse Grossccb13522011-10-25 19:26:31 -0700618 } else if (key->eth.type == htons(ETH_P_IPV6)) {
619 int nh_len; /* IPv6 Header + Extensions */
620
Andy Zhou03f0d912013-08-07 20:01:00 -0700621 nh_len = parse_ipv6hdr(skb, key);
Jesse Grossccb13522011-10-25 19:26:31 -0700622 if (unlikely(nh_len < 0)) {
Jesse Gross07148122014-10-03 15:35:29 -0700623 memset(&key->ip, 0, sizeof(key->ip));
624 memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr));
Andy Zhou03f0d912013-08-07 20:01:00 -0700625 if (nh_len == -EINVAL) {
Jesse Grossccb13522011-10-25 19:26:31 -0700626 skb->transport_header = skb->network_header;
Andy Zhou03f0d912013-08-07 20:01:00 -0700627 error = 0;
628 } else {
Jesse Grossccb13522011-10-25 19:26:31 -0700629 error = nh_len;
Andy Zhou03f0d912013-08-07 20:01:00 -0700630 }
631 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700632 }
633
634 if (key->ip.frag == OVS_FRAG_TYPE_LATER)
Andy Zhou03f0d912013-08-07 20:01:00 -0700635 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700636 if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
637 key->ip.frag = OVS_FRAG_TYPE_FIRST;
638
639 /* Transport layer. */
640 if (key->ip.proto == NEXTHDR_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700641 if (tcphdr_ok(skb)) {
642 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700643 key->tp.src = tcp->source;
644 key->tp.dst = tcp->dest;
645 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700646 } else {
647 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700648 }
649 } else if (key->ip.proto == NEXTHDR_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700650 if (udphdr_ok(skb)) {
651 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700652 key->tp.src = udp->source;
653 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700654 } else {
655 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700656 }
Joe Stringera175a722013-08-22 12:30:48 -0700657 } else if (key->ip.proto == NEXTHDR_SCTP) {
658 if (sctphdr_ok(skb)) {
659 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700660 key->tp.src = sctp->source;
661 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700662 } else {
663 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700664 }
Jesse Grossccb13522011-10-25 19:26:31 -0700665 } else if (key->ip.proto == NEXTHDR_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700666 if (icmp6hdr_ok(skb)) {
Andy Zhou03f0d912013-08-07 20:01:00 -0700667 error = parse_icmpv6(skb, key, nh_len);
668 if (error)
669 return error;
Jesse Gross07148122014-10-03 15:35:29 -0700670 } else {
671 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700672 }
673 }
674 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700675 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700676}
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700677
Andy Zhou971427f32014-09-15 19:37:25 -0700678int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key)
679{
680 return key_extract(skb, key);
681}
682
Thomas Graf12eb18f2014-11-06 06:58:52 -0800683int ovs_flow_key_extract(const struct ovs_tunnel_info *tun_info,
Pravin B Shelar8c8b1b82014-09-15 19:28:44 -0700684 struct sk_buff *skb, struct sw_flow_key *key)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700685{
686 /* Extract metadata from packet. */
Jesse Grossf5796682014-10-03 15:35:33 -0700687 if (tun_info) {
Jesse Grossf0b128c2014-10-03 15:35:31 -0700688 memcpy(&key->tun_key, &tun_info->tunnel, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700689
690 if (tun_info->options) {
691 BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
692 8)) - 1
693 > sizeof(key->tun_opts));
Thomas Grafd91641d2015-01-15 03:53:57 +0100694 memcpy(TUN_METADATA_OPTS(key, tun_info->options_len),
Jesse Grossf5796682014-10-03 15:35:33 -0700695 tun_info->options, tun_info->options_len);
696 key->tun_opts_len = tun_info->options_len;
697 } else {
698 key->tun_opts_len = 0;
699 }
700 } else {
701 key->tun_opts_len = 0;
Jesse Gross07148122014-10-03 15:35:29 -0700702 memset(&key->tun_key, 0, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700703 }
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700704
705 key->phy.priority = skb->priority;
706 key->phy.in_port = OVS_CB(skb)->input_vport->port_no;
707 key->phy.skb_mark = skb->mark;
Jesse Gross07148122014-10-03 15:35:29 -0700708 key->ovs_flow_hash = 0;
709 key->recirc_id = 0;
710
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700711 return key_extract(skb, key);
712}
713
714int ovs_flow_key_extract_userspace(const struct nlattr *attr,
715 struct sk_buff *skb,
Jarno Rajahalme05da5892014-11-06 07:03:05 -0800716 struct sw_flow_key *key, bool log)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700717{
718 int err;
719
Pravin B Shelarb35725a2015-02-10 13:35:16 -0800720 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
721
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700722 /* Extract metadata from netlink attributes. */
Jarno Rajahalme05da5892014-11-06 07:03:05 -0800723 err = ovs_nla_get_flow_metadata(attr, key, log);
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700724 if (err)
725 return err;
726
727 return key_extract(skb, key);
728}