blob: 9db87331e2111144b223d831045715d928ce4d2f [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"
Joe Stringer7f8a4362015-08-26 11:31:48 -070052#include "conntrack.h"
Pravin B Shelar83c8df22014-09-15 19:20:31 -070053
Pravin B Shelare6445712013-10-03 18:16:47 -070054u64 ovs_flow_used_time(unsigned long flow_jiffies)
Andy Zhou03f0d912013-08-07 20:01:00 -070055{
Pravin B Shelare6445712013-10-03 18:16:47 -070056 struct timespec cur_ts;
57 u64 cur_ms, idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070058
Pravin B Shelare6445712013-10-03 18:16:47 -070059 ktime_get_ts(&cur_ts);
60 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
61 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
62 cur_ts.tv_nsec / NSEC_PER_MSEC;
Andy Zhou03f0d912013-08-07 20:01:00 -070063
Pravin B Shelare6445712013-10-03 18:16:47 -070064 return cur_ms - idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070065}
66
Jarno Rajahalmedf23e9f2013-10-23 01:40:44 -070067#define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
Andy Zhou03f0d912013-08-07 20:01:00 -070068
Ben Pfaffad552002014-05-06 16:48:38 -070069void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags,
Thomas Graf12eb18f2014-11-06 06:58:52 -080070 const struct sk_buff *skb)
Andy Zhou5828cd92013-08-27 13:02:21 -070071{
Pravin B Shelare298e502013-10-29 17:22:21 -070072 struct flow_stats *stats;
Jarno Rajahalme63e79592014-03-27 12:42:54 -070073 int node = numa_node_id();
Jiri Pirkodf8a39d2015-01-13 17:13:44 +010074 int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Andy Zhou5828cd92013-08-27 13:02:21 -070075
Jarno Rajahalme63e79592014-03-27 12:42:54 -070076 stats = rcu_dereference(flow->stats[node]);
Pravin B Shelare298e502013-10-29 17:22:21 -070077
Jarno Rajahalme63e79592014-03-27 12:42:54 -070078 /* Check if already have node-specific stats. */
79 if (likely(stats)) {
80 spin_lock(&stats->lock);
81 /* Mark if we write on the pre-allocated stats. */
82 if (node == 0 && unlikely(flow->stats_last_writer != node))
83 flow->stats_last_writer = node;
84 } else {
85 stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */
86 spin_lock(&stats->lock);
87
88 /* If the current NUMA-node is the only writer on the
89 * pre-allocated stats keep using them.
90 */
91 if (unlikely(flow->stats_last_writer != node)) {
92 /* A previous locker may have already allocated the
93 * stats, so we need to check again. If node-specific
94 * stats were already allocated, we update the pre-
95 * allocated stats as we have already locked them.
96 */
97 if (likely(flow->stats_last_writer != NUMA_NO_NODE)
Andreea-Cristina Bernat8c6b00c2014-08-17 16:29:43 +030098 && likely(!rcu_access_pointer(flow->stats[node]))) {
Jarno Rajahalme63e79592014-03-27 12:42:54 -070099 /* Try to allocate node-specific stats. */
100 struct flow_stats *new_stats;
101
102 new_stats =
103 kmem_cache_alloc_node(flow_stats_cache,
David Rientjes4167e9b2015-04-14 15:46:55 -0700104 GFP_NOWAIT |
105 __GFP_THISNODE |
106 __GFP_NOWARN |
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700107 __GFP_NOMEMALLOC,
108 node);
109 if (likely(new_stats)) {
110 new_stats->used = jiffies;
111 new_stats->packet_count = 1;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800112 new_stats->byte_count = len;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700113 new_stats->tcp_flags = tcp_flags;
114 spin_lock_init(&new_stats->lock);
115
116 rcu_assign_pointer(flow->stats[node],
117 new_stats);
118 goto unlock;
119 }
120 }
121 flow->stats_last_writer = node;
122 }
123 }
124
Pravin B Shelare298e502013-10-29 17:22:21 -0700125 stats->used = jiffies;
126 stats->packet_count++;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800127 stats->byte_count += len;
Pravin B Shelare298e502013-10-29 17:22:21 -0700128 stats->tcp_flags |= tcp_flags;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700129unlock:
Pravin B Shelare298e502013-10-29 17:22:21 -0700130 spin_unlock(&stats->lock);
131}
132
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700133/* Must be called with rcu_read_lock or ovs_mutex. */
134void ovs_flow_stats_get(const struct sw_flow *flow,
135 struct ovs_flow_stats *ovs_stats,
Pravin B Shelare298e502013-10-29 17:22:21 -0700136 unsigned long *used, __be16 *tcp_flags)
137{
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700138 int node;
Pravin B Shelare298e502013-10-29 17:22:21 -0700139
140 *used = 0;
141 *tcp_flags = 0;
142 memset(ovs_stats, 0, sizeof(*ovs_stats));
143
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700144 for_each_node(node) {
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700145 struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[node]);
Pravin B Shelare298e502013-10-29 17:22:21 -0700146
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700147 if (stats) {
148 /* Local CPU may write on non-local stats, so we must
149 * block bottom-halves here.
150 */
151 spin_lock_bh(&stats->lock);
152 if (!*used || time_after(stats->used, *used))
153 *used = stats->used;
154 *tcp_flags |= stats->tcp_flags;
155 ovs_stats->n_packets += stats->packet_count;
156 ovs_stats->n_bytes += stats->byte_count;
157 spin_unlock_bh(&stats->lock);
158 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700159 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700160}
161
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700162/* Called with ovs_mutex. */
Pravin B Shelare298e502013-10-29 17:22:21 -0700163void ovs_flow_stats_clear(struct sw_flow *flow)
164{
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700165 int node;
Pravin B Shelare298e502013-10-29 17:22:21 -0700166
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700167 for_each_node(node) {
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700168 struct flow_stats *stats = ovsl_dereference(flow->stats[node]);
Jarno Rajahalme23dabf82014-03-27 12:35:23 -0700169
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700170 if (stats) {
171 spin_lock_bh(&stats->lock);
172 stats->used = 0;
173 stats->packet_count = 0;
174 stats->byte_count = 0;
175 stats->tcp_flags = 0;
176 spin_unlock_bh(&stats->lock);
177 }
178 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700179}
180
Jesse Grossccb13522011-10-25 19:26:31 -0700181static int check_header(struct sk_buff *skb, int len)
182{
183 if (unlikely(skb->len < len))
184 return -EINVAL;
185 if (unlikely(!pskb_may_pull(skb, len)))
186 return -ENOMEM;
187 return 0;
188}
189
190static bool arphdr_ok(struct sk_buff *skb)
191{
192 return pskb_may_pull(skb, skb_network_offset(skb) +
193 sizeof(struct arp_eth_header));
194}
195
196static int check_iphdr(struct sk_buff *skb)
197{
198 unsigned int nh_ofs = skb_network_offset(skb);
199 unsigned int ip_len;
200 int err;
201
202 err = check_header(skb, nh_ofs + sizeof(struct iphdr));
203 if (unlikely(err))
204 return err;
205
206 ip_len = ip_hdrlen(skb);
207 if (unlikely(ip_len < sizeof(struct iphdr) ||
208 skb->len < nh_ofs + ip_len))
209 return -EINVAL;
210
211 skb_set_transport_header(skb, nh_ofs + ip_len);
212 return 0;
213}
214
215static bool tcphdr_ok(struct sk_buff *skb)
216{
217 int th_ofs = skb_transport_offset(skb);
218 int tcp_len;
219
220 if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
221 return false;
222
223 tcp_len = tcp_hdrlen(skb);
224 if (unlikely(tcp_len < sizeof(struct tcphdr) ||
225 skb->len < th_ofs + tcp_len))
226 return false;
227
228 return true;
229}
230
231static bool udphdr_ok(struct sk_buff *skb)
232{
233 return pskb_may_pull(skb, skb_transport_offset(skb) +
234 sizeof(struct udphdr));
235}
236
Joe Stringera175a722013-08-22 12:30:48 -0700237static bool sctphdr_ok(struct sk_buff *skb)
238{
239 return pskb_may_pull(skb, skb_transport_offset(skb) +
240 sizeof(struct sctphdr));
241}
242
Jesse Grossccb13522011-10-25 19:26:31 -0700243static bool icmphdr_ok(struct sk_buff *skb)
244{
245 return pskb_may_pull(skb, skb_transport_offset(skb) +
246 sizeof(struct icmphdr));
247}
248
Andy Zhou03f0d912013-08-07 20:01:00 -0700249static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700250{
251 unsigned int nh_ofs = skb_network_offset(skb);
252 unsigned int nh_len;
253 int payload_ofs;
254 struct ipv6hdr *nh;
255 uint8_t nexthdr;
256 __be16 frag_off;
257 int err;
258
Jesse Grossccb13522011-10-25 19:26:31 -0700259 err = check_header(skb, nh_ofs + sizeof(*nh));
260 if (unlikely(err))
261 return err;
262
263 nh = ipv6_hdr(skb);
264 nexthdr = nh->nexthdr;
265 payload_ofs = (u8 *)(nh + 1) - skb->data;
266
267 key->ip.proto = NEXTHDR_NONE;
268 key->ip.tos = ipv6_get_dsfield(nh);
269 key->ip.ttl = nh->hop_limit;
270 key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
271 key->ipv6.addr.src = nh->saddr;
272 key->ipv6.addr.dst = nh->daddr;
273
274 payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off);
Jesse Grossccb13522011-10-25 19:26:31 -0700275
276 if (frag_off) {
277 if (frag_off & htons(~0x7))
278 key->ip.frag = OVS_FRAG_TYPE_LATER;
279 else
280 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700281 } else {
282 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700283 }
284
Simon Hormanc30da492015-08-29 09:02:21 +0900285 /* Delayed handling of error in ipv6_skip_exthdr() as it
286 * always sets frag_off to a valid value which may be
287 * used to set key->ip.frag above.
288 */
289 if (unlikely(payload_ofs < 0))
290 return -EPROTO;
291
Jesse Grossccb13522011-10-25 19:26:31 -0700292 nh_len = payload_ofs - nh_ofs;
293 skb_set_transport_header(skb, nh_ofs + nh_len);
294 key->ip.proto = nexthdr;
295 return nh_len;
296}
297
298static bool icmp6hdr_ok(struct sk_buff *skb)
299{
300 return pskb_may_pull(skb, skb_transport_offset(skb) +
301 sizeof(struct icmp6hdr));
302}
303
Jesse Grossccb13522011-10-25 19:26:31 -0700304static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
305{
306 struct qtag_prefix {
307 __be16 eth_type; /* ETH_P_8021Q */
308 __be16 tci;
309 };
310 struct qtag_prefix *qp;
311
312 if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)))
313 return 0;
314
315 if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) +
316 sizeof(__be16))))
317 return -ENOMEM;
318
319 qp = (struct qtag_prefix *) skb->data;
320 key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT);
321 __skb_pull(skb, sizeof(struct qtag_prefix));
322
323 return 0;
324}
325
326static __be16 parse_ethertype(struct sk_buff *skb)
327{
328 struct llc_snap_hdr {
329 u8 dsap; /* Always 0xAA */
330 u8 ssap; /* Always 0xAA */
331 u8 ctrl;
332 u8 oui[3];
333 __be16 ethertype;
334 };
335 struct llc_snap_hdr *llc;
336 __be16 proto;
337
338 proto = *(__be16 *) skb->data;
339 __skb_pull(skb, sizeof(__be16));
340
Alexander Duyck6713fc92015-05-04 14:34:05 -0700341 if (eth_proto_is_802_3(proto))
Jesse Grossccb13522011-10-25 19:26:31 -0700342 return proto;
343
344 if (skb->len < sizeof(struct llc_snap_hdr))
345 return htons(ETH_P_802_2);
346
347 if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
348 return htons(0);
349
350 llc = (struct llc_snap_hdr *) skb->data;
351 if (llc->dsap != LLC_SAP_SNAP ||
352 llc->ssap != LLC_SAP_SNAP ||
353 (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
354 return htons(ETH_P_802_2);
355
356 __skb_pull(skb, sizeof(struct llc_snap_hdr));
Rich Lane17b682a2013-02-19 11:10:30 -0800357
Alexander Duyck6713fc92015-05-04 14:34:05 -0700358 if (eth_proto_is_802_3(llc->ethertype))
Rich Lane17b682a2013-02-19 11:10:30 -0800359 return llc->ethertype;
360
361 return htons(ETH_P_802_2);
Jesse Grossccb13522011-10-25 19:26:31 -0700362}
363
364static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
Andy Zhou03f0d912013-08-07 20:01:00 -0700365 int nh_len)
Jesse Grossccb13522011-10-25 19:26:31 -0700366{
367 struct icmp6hdr *icmp = icmp6_hdr(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700368
369 /* The ICMPv6 type and code fields use the 16-bit transport port
370 * fields, so we need to store them in 16-bit network byte order.
371 */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700372 key->tp.src = htons(icmp->icmp6_type);
373 key->tp.dst = htons(icmp->icmp6_code);
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700374 memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd));
Jesse Grossccb13522011-10-25 19:26:31 -0700375
376 if (icmp->icmp6_code == 0 &&
377 (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
378 icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
379 int icmp_len = skb->len - skb_transport_offset(skb);
380 struct nd_msg *nd;
381 int offset;
382
Jesse Grossccb13522011-10-25 19:26:31 -0700383 /* In order to process neighbor discovery options, we need the
384 * entire packet.
385 */
386 if (unlikely(icmp_len < sizeof(*nd)))
Andy Zhou03f0d912013-08-07 20:01:00 -0700387 return 0;
388
389 if (unlikely(skb_linearize(skb)))
390 return -ENOMEM;
Jesse Grossccb13522011-10-25 19:26:31 -0700391
392 nd = (struct nd_msg *)skb_transport_header(skb);
393 key->ipv6.nd.target = nd->target;
Jesse Grossccb13522011-10-25 19:26:31 -0700394
395 icmp_len -= sizeof(*nd);
396 offset = 0;
397 while (icmp_len >= 8) {
398 struct nd_opt_hdr *nd_opt =
399 (struct nd_opt_hdr *)(nd->opt + offset);
400 int opt_len = nd_opt->nd_opt_len * 8;
401
402 if (unlikely(!opt_len || opt_len > icmp_len))
Andy Zhou03f0d912013-08-07 20:01:00 -0700403 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700404
405 /* Store the link layer address if the appropriate
406 * option is provided. It is considered an error if
407 * the same link layer option is specified twice.
408 */
409 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
410 && opt_len == 8) {
411 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
412 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800413 ether_addr_copy(key->ipv6.nd.sll,
414 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700415 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
416 && opt_len == 8) {
417 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
418 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800419 ether_addr_copy(key->ipv6.nd.tll,
420 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700421 }
422
423 icmp_len -= opt_len;
424 offset += opt_len;
425 }
426 }
427
Andy Zhou03f0d912013-08-07 20:01:00 -0700428 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700429
430invalid:
431 memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
432 memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
433 memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
434
Andy Zhou03f0d912013-08-07 20:01:00 -0700435 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700436}
437
438/**
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700439 * key_extract - extracts a flow key from an Ethernet frame.
Jesse Grossccb13522011-10-25 19:26:31 -0700440 * @skb: sk_buff that contains the frame, with skb->data pointing to the
441 * Ethernet header
Jesse Grossccb13522011-10-25 19:26:31 -0700442 * @key: output flow key
Jesse Grossccb13522011-10-25 19:26:31 -0700443 *
444 * The caller must ensure that skb->len >= ETH_HLEN.
445 *
446 * Returns 0 if successful, otherwise a negative errno value.
447 *
448 * Initializes @skb header pointers as follows:
449 *
450 * - skb->mac_header: the Ethernet header.
451 *
452 * - skb->network_header: just past the Ethernet header, or just past the
453 * VLAN header, to the first byte of the Ethernet payload.
454 *
Lorand Jakab34d94f22013-06-03 10:01:14 -0700455 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
Jesse Grossccb13522011-10-25 19:26:31 -0700456 * on output, then just past the IP header, if one is present and
457 * of a correct length, otherwise the same as skb->network_header.
Lorand Jakab34d94f22013-06-03 10:01:14 -0700458 * For other key->eth.type values it is left untouched.
Jesse Grossccb13522011-10-25 19:26:31 -0700459 */
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700460static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700461{
Andy Zhou03f0d912013-08-07 20:01:00 -0700462 int error;
Jesse Grossccb13522011-10-25 19:26:31 -0700463 struct ethhdr *eth;
464
Jesse Grossf5796682014-10-03 15:35:33 -0700465 /* Flags are always used as part of stats */
466 key->tp.flags = 0;
467
Jesse Grossccb13522011-10-25 19:26:31 -0700468 skb_reset_mac_header(skb);
469
470 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
471 * header in the linear data area.
472 */
473 eth = eth_hdr(skb);
Joe Perches8c63ff02014-02-18 11:15:45 -0800474 ether_addr_copy(key->eth.src, eth->h_source);
475 ether_addr_copy(key->eth.dst, eth->h_dest);
Jesse Grossccb13522011-10-25 19:26:31 -0700476
477 __skb_pull(skb, 2 * ETH_ALEN);
Pravin B Shelarb34df5e2013-06-13 11:11:44 -0700478 /* We are going to push all headers that we pull, so no need to
479 * update skb->csum here.
480 */
Jesse Grossccb13522011-10-25 19:26:31 -0700481
Jesse Gross07148122014-10-03 15:35:29 -0700482 key->eth.tci = 0;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +0100483 if (skb_vlan_tag_present(skb))
Jesse Grossccb13522011-10-25 19:26:31 -0700484 key->eth.tci = htons(skb->vlan_tci);
485 else if (eth->h_proto == htons(ETH_P_8021Q))
486 if (unlikely(parse_vlan(skb, key)))
487 return -ENOMEM;
488
489 key->eth.type = parse_ethertype(skb);
490 if (unlikely(key->eth.type == htons(0)))
491 return -ENOMEM;
492
493 skb_reset_network_header(skb);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700494 skb_reset_mac_len(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700495 __skb_push(skb, skb->data - skb_mac_header(skb));
496
497 /* Network layer. */
498 if (key->eth.type == htons(ETH_P_IP)) {
499 struct iphdr *nh;
500 __be16 offset;
501
Jesse Grossccb13522011-10-25 19:26:31 -0700502 error = check_iphdr(skb);
503 if (unlikely(error)) {
Jesse Gross07148122014-10-03 15:35:29 -0700504 memset(&key->ip, 0, sizeof(key->ip));
505 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700506 if (error == -EINVAL) {
507 skb->transport_header = skb->network_header;
508 error = 0;
509 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700510 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700511 }
512
513 nh = ip_hdr(skb);
514 key->ipv4.addr.src = nh->saddr;
515 key->ipv4.addr.dst = nh->daddr;
516
517 key->ip.proto = nh->protocol;
518 key->ip.tos = nh->tos;
519 key->ip.ttl = nh->ttl;
520
521 offset = nh->frag_off & htons(IP_OFFSET);
522 if (offset) {
523 key->ip.frag = OVS_FRAG_TYPE_LATER;
Andy Zhou03f0d912013-08-07 20:01:00 -0700524 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700525 }
526 if (nh->frag_off & htons(IP_MF) ||
Jesse Gross07148122014-10-03 15:35:29 -0700527 skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
Jesse Grossccb13522011-10-25 19:26:31 -0700528 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Jesse Gross07148122014-10-03 15:35:29 -0700529 else
530 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700531
532 /* Transport layer. */
533 if (key->ip.proto == IPPROTO_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700534 if (tcphdr_ok(skb)) {
535 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700536 key->tp.src = tcp->source;
537 key->tp.dst = tcp->dest;
538 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700539 } else {
540 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700541 }
Jesse Gross07148122014-10-03 15:35:29 -0700542
Jesse Grossccb13522011-10-25 19:26:31 -0700543 } else if (key->ip.proto == IPPROTO_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700544 if (udphdr_ok(skb)) {
545 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700546 key->tp.src = udp->source;
547 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700548 } else {
549 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700550 }
Joe Stringera175a722013-08-22 12:30:48 -0700551 } else if (key->ip.proto == IPPROTO_SCTP) {
552 if (sctphdr_ok(skb)) {
553 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700554 key->tp.src = sctp->source;
555 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700556 } else {
557 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700558 }
Jesse Grossccb13522011-10-25 19:26:31 -0700559 } else if (key->ip.proto == IPPROTO_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700560 if (icmphdr_ok(skb)) {
561 struct icmphdr *icmp = icmp_hdr(skb);
562 /* The ICMP type and code fields use the 16-bit
563 * transport port fields, so we need to store
564 * them in 16-bit network byte order. */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700565 key->tp.src = htons(icmp->type);
566 key->tp.dst = htons(icmp->code);
Jesse Gross07148122014-10-03 15:35:29 -0700567 } else {
568 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700569 }
570 }
571
Jesse Gross07148122014-10-03 15:35:29 -0700572 } else if (key->eth.type == htons(ETH_P_ARP) ||
573 key->eth.type == htons(ETH_P_RARP)) {
Jesse Grossccb13522011-10-25 19:26:31 -0700574 struct arp_eth_header *arp;
Li RongQing389f4892014-10-17 14:03:08 +0800575 bool arp_available = arphdr_ok(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700576
577 arp = (struct arp_eth_header *)skb_network_header(skb);
578
Li RongQing389f4892014-10-17 14:03:08 +0800579 if (arp_available &&
Jesse Gross07148122014-10-03 15:35:29 -0700580 arp->ar_hrd == htons(ARPHRD_ETHER) &&
581 arp->ar_pro == htons(ETH_P_IP) &&
582 arp->ar_hln == ETH_ALEN &&
583 arp->ar_pln == 4) {
Jesse Grossccb13522011-10-25 19:26:31 -0700584
585 /* We only match on the lower 8 bits of the opcode. */
586 if (ntohs(arp->ar_op) <= 0xff)
587 key->ip.proto = ntohs(arp->ar_op);
Jesse Gross07148122014-10-03 15:35:29 -0700588 else
589 key->ip.proto = 0;
590
Mehak Mahajand04d3822012-10-30 15:50:28 -0700591 memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
592 memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
Joe Perches8c63ff02014-02-18 11:15:45 -0800593 ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
594 ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
Jesse Gross07148122014-10-03 15:35:29 -0700595 } else {
596 memset(&key->ip, 0, sizeof(key->ip));
597 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700598 }
Simon Horman25cd9ba2014-10-06 05:05:13 -0700599 } else if (eth_p_mpls(key->eth.type)) {
600 size_t stack_len = MPLS_HLEN;
601
602 /* In the presence of an MPLS label stack the end of the L2
603 * header and the beginning of the L3 header differ.
604 *
605 * Advance network_header to the beginning of the L3
606 * header. mac_len corresponds to the end of the L2 header.
607 */
608 while (1) {
609 __be32 lse;
610
611 error = check_header(skb, skb->mac_len + stack_len);
612 if (unlikely(error))
613 return 0;
614
615 memcpy(&lse, skb_network_header(skb), MPLS_HLEN);
616
617 if (stack_len == MPLS_HLEN)
618 memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN);
619
620 skb_set_network_header(skb, skb->mac_len + stack_len);
621 if (lse & htonl(MPLS_LS_S_MASK))
622 break;
623
624 stack_len += MPLS_HLEN;
625 }
Jesse Grossccb13522011-10-25 19:26:31 -0700626 } else if (key->eth.type == htons(ETH_P_IPV6)) {
627 int nh_len; /* IPv6 Header + Extensions */
628
Andy Zhou03f0d912013-08-07 20:01:00 -0700629 nh_len = parse_ipv6hdr(skb, key);
Jesse Grossccb13522011-10-25 19:26:31 -0700630 if (unlikely(nh_len < 0)) {
Simon Hormanc30da492015-08-29 09:02:21 +0900631 switch (nh_len) {
632 case -EINVAL:
633 memset(&key->ip, 0, sizeof(key->ip));
634 memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr));
635 /* fall-through */
636 case -EPROTO:
Jesse Grossccb13522011-10-25 19:26:31 -0700637 skb->transport_header = skb->network_header;
Andy Zhou03f0d912013-08-07 20:01:00 -0700638 error = 0;
Simon Hormanc30da492015-08-29 09:02:21 +0900639 break;
640 default:
Jesse Grossccb13522011-10-25 19:26:31 -0700641 error = nh_len;
Andy Zhou03f0d912013-08-07 20:01:00 -0700642 }
643 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700644 }
645
646 if (key->ip.frag == OVS_FRAG_TYPE_LATER)
Andy Zhou03f0d912013-08-07 20:01:00 -0700647 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700648 if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
649 key->ip.frag = OVS_FRAG_TYPE_FIRST;
650
651 /* Transport layer. */
652 if (key->ip.proto == NEXTHDR_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700653 if (tcphdr_ok(skb)) {
654 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700655 key->tp.src = tcp->source;
656 key->tp.dst = tcp->dest;
657 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700658 } else {
659 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700660 }
661 } else if (key->ip.proto == NEXTHDR_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700662 if (udphdr_ok(skb)) {
663 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700664 key->tp.src = udp->source;
665 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700666 } else {
667 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700668 }
Joe Stringera175a722013-08-22 12:30:48 -0700669 } else if (key->ip.proto == NEXTHDR_SCTP) {
670 if (sctphdr_ok(skb)) {
671 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700672 key->tp.src = sctp->source;
673 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700674 } else {
675 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700676 }
Jesse Grossccb13522011-10-25 19:26:31 -0700677 } else if (key->ip.proto == NEXTHDR_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700678 if (icmp6hdr_ok(skb)) {
Andy Zhou03f0d912013-08-07 20:01:00 -0700679 error = parse_icmpv6(skb, key, nh_len);
680 if (error)
681 return error;
Jesse Gross07148122014-10-03 15:35:29 -0700682 } else {
683 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700684 }
685 }
686 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700687 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700688}
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700689
Andy Zhou971427f32014-09-15 19:37:25 -0700690int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key)
691{
692 return key_extract(skb, key);
693}
694
Thomas Graf1d8fff92015-07-21 10:43:54 +0200695int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
Pravin B Shelar8c8b1b82014-09-15 19:28:44 -0700696 struct sk_buff *skb, struct sw_flow_key *key)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700697{
698 /* Extract metadata from packet. */
Jesse Grossf5796682014-10-03 15:35:33 -0700699 if (tun_info) {
Jiri Benc7f9562a2015-08-28 20:48:20 +0200700 if (ip_tunnel_info_af(tun_info) != AF_INET)
701 return -EINVAL;
Thomas Graf1d8fff92015-07-21 10:43:54 +0200702 memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700703
704 if (tun_info->options) {
705 BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
706 8)) - 1
707 > sizeof(key->tun_opts));
Thomas Grafd91641d2015-01-15 03:53:57 +0100708 memcpy(TUN_METADATA_OPTS(key, tun_info->options_len),
Jesse Grossf5796682014-10-03 15:35:33 -0700709 tun_info->options, tun_info->options_len);
710 key->tun_opts_len = tun_info->options_len;
711 } else {
712 key->tun_opts_len = 0;
713 }
714 } else {
715 key->tun_opts_len = 0;
Jesse Gross07148122014-10-03 15:35:29 -0700716 memset(&key->tun_key, 0, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700717 }
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700718
719 key->phy.priority = skb->priority;
720 key->phy.in_port = OVS_CB(skb)->input_vport->port_no;
721 key->phy.skb_mark = skb->mark;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700722 ovs_ct_fill_key(skb, key);
Jesse Gross07148122014-10-03 15:35:29 -0700723 key->ovs_flow_hash = 0;
724 key->recirc_id = 0;
725
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700726 return key_extract(skb, key);
727}
728
Joe Stringerc2ac6672015-08-26 11:31:52 -0700729int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700730 struct sk_buff *skb,
Jarno Rajahalme05da5892014-11-06 07:03:05 -0800731 struct sw_flow_key *key, bool log)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700732{
733 int err;
734
Pravin B Shelarb35725a2015-02-10 13:35:16 -0800735 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
736
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700737 /* Extract metadata from netlink attributes. */
Joe Stringerc2ac6672015-08-26 11:31:52 -0700738 err = ovs_nla_get_flow_metadata(net, attr, key, log);
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700739 if (err)
740 return err;
741
742 return key_extract(skb, key);
743}