blob: 5a3195e538ce588ca6889bce445f61d02cf59894 [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);
275 if (unlikely(payload_ofs < 0))
276 return -EINVAL;
277
278 if (frag_off) {
279 if (frag_off & htons(~0x7))
280 key->ip.frag = OVS_FRAG_TYPE_LATER;
281 else
282 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700283 } else {
284 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700285 }
286
287 nh_len = payload_ofs - nh_ofs;
288 skb_set_transport_header(skb, nh_ofs + nh_len);
289 key->ip.proto = nexthdr;
290 return nh_len;
291}
292
293static bool icmp6hdr_ok(struct sk_buff *skb)
294{
295 return pskb_may_pull(skb, skb_transport_offset(skb) +
296 sizeof(struct icmp6hdr));
297}
298
Jesse Grossccb13522011-10-25 19:26:31 -0700299static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
300{
301 struct qtag_prefix {
302 __be16 eth_type; /* ETH_P_8021Q */
303 __be16 tci;
304 };
305 struct qtag_prefix *qp;
306
307 if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16)))
308 return 0;
309
310 if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) +
311 sizeof(__be16))))
312 return -ENOMEM;
313
314 qp = (struct qtag_prefix *) skb->data;
315 key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT);
316 __skb_pull(skb, sizeof(struct qtag_prefix));
317
318 return 0;
319}
320
321static __be16 parse_ethertype(struct sk_buff *skb)
322{
323 struct llc_snap_hdr {
324 u8 dsap; /* Always 0xAA */
325 u8 ssap; /* Always 0xAA */
326 u8 ctrl;
327 u8 oui[3];
328 __be16 ethertype;
329 };
330 struct llc_snap_hdr *llc;
331 __be16 proto;
332
333 proto = *(__be16 *) skb->data;
334 __skb_pull(skb, sizeof(__be16));
335
Alexander Duyck6713fc92015-05-04 14:34:05 -0700336 if (eth_proto_is_802_3(proto))
Jesse Grossccb13522011-10-25 19:26:31 -0700337 return proto;
338
339 if (skb->len < sizeof(struct llc_snap_hdr))
340 return htons(ETH_P_802_2);
341
342 if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
343 return htons(0);
344
345 llc = (struct llc_snap_hdr *) skb->data;
346 if (llc->dsap != LLC_SAP_SNAP ||
347 llc->ssap != LLC_SAP_SNAP ||
348 (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
349 return htons(ETH_P_802_2);
350
351 __skb_pull(skb, sizeof(struct llc_snap_hdr));
Rich Lane17b682a2013-02-19 11:10:30 -0800352
Alexander Duyck6713fc92015-05-04 14:34:05 -0700353 if (eth_proto_is_802_3(llc->ethertype))
Rich Lane17b682a2013-02-19 11:10:30 -0800354 return llc->ethertype;
355
356 return htons(ETH_P_802_2);
Jesse Grossccb13522011-10-25 19:26:31 -0700357}
358
359static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
Andy Zhou03f0d912013-08-07 20:01:00 -0700360 int nh_len)
Jesse Grossccb13522011-10-25 19:26:31 -0700361{
362 struct icmp6hdr *icmp = icmp6_hdr(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700363
364 /* The ICMPv6 type and code fields use the 16-bit transport port
365 * fields, so we need to store them in 16-bit network byte order.
366 */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700367 key->tp.src = htons(icmp->icmp6_type);
368 key->tp.dst = htons(icmp->icmp6_code);
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700369 memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd));
Jesse Grossccb13522011-10-25 19:26:31 -0700370
371 if (icmp->icmp6_code == 0 &&
372 (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
373 icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
374 int icmp_len = skb->len - skb_transport_offset(skb);
375 struct nd_msg *nd;
376 int offset;
377
Jesse Grossccb13522011-10-25 19:26:31 -0700378 /* In order to process neighbor discovery options, we need the
379 * entire packet.
380 */
381 if (unlikely(icmp_len < sizeof(*nd)))
Andy Zhou03f0d912013-08-07 20:01:00 -0700382 return 0;
383
384 if (unlikely(skb_linearize(skb)))
385 return -ENOMEM;
Jesse Grossccb13522011-10-25 19:26:31 -0700386
387 nd = (struct nd_msg *)skb_transport_header(skb);
388 key->ipv6.nd.target = nd->target;
Jesse Grossccb13522011-10-25 19:26:31 -0700389
390 icmp_len -= sizeof(*nd);
391 offset = 0;
392 while (icmp_len >= 8) {
393 struct nd_opt_hdr *nd_opt =
394 (struct nd_opt_hdr *)(nd->opt + offset);
395 int opt_len = nd_opt->nd_opt_len * 8;
396
397 if (unlikely(!opt_len || opt_len > icmp_len))
Andy Zhou03f0d912013-08-07 20:01:00 -0700398 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700399
400 /* Store the link layer address if the appropriate
401 * option is provided. It is considered an error if
402 * the same link layer option is specified twice.
403 */
404 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
405 && opt_len == 8) {
406 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
407 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800408 ether_addr_copy(key->ipv6.nd.sll,
409 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700410 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
411 && opt_len == 8) {
412 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
413 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800414 ether_addr_copy(key->ipv6.nd.tll,
415 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700416 }
417
418 icmp_len -= opt_len;
419 offset += opt_len;
420 }
421 }
422
Andy Zhou03f0d912013-08-07 20:01:00 -0700423 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700424
425invalid:
426 memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
427 memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
428 memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
429
Andy Zhou03f0d912013-08-07 20:01:00 -0700430 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700431}
432
433/**
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700434 * key_extract - extracts a flow key from an Ethernet frame.
Jesse Grossccb13522011-10-25 19:26:31 -0700435 * @skb: sk_buff that contains the frame, with skb->data pointing to the
436 * Ethernet header
Jesse Grossccb13522011-10-25 19:26:31 -0700437 * @key: output flow key
Jesse Grossccb13522011-10-25 19:26:31 -0700438 *
439 * The caller must ensure that skb->len >= ETH_HLEN.
440 *
441 * Returns 0 if successful, otherwise a negative errno value.
442 *
443 * Initializes @skb header pointers as follows:
444 *
445 * - skb->mac_header: the Ethernet header.
446 *
447 * - skb->network_header: just past the Ethernet header, or just past the
448 * VLAN header, to the first byte of the Ethernet payload.
449 *
Lorand Jakab34d94f22013-06-03 10:01:14 -0700450 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
Jesse Grossccb13522011-10-25 19:26:31 -0700451 * on output, then just past the IP header, if one is present and
452 * of a correct length, otherwise the same as skb->network_header.
Lorand Jakab34d94f22013-06-03 10:01:14 -0700453 * For other key->eth.type values it is left untouched.
Jesse Grossccb13522011-10-25 19:26:31 -0700454 */
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700455static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700456{
Andy Zhou03f0d912013-08-07 20:01:00 -0700457 int error;
Jesse Grossccb13522011-10-25 19:26:31 -0700458 struct ethhdr *eth;
459
Jesse Grossf5796682014-10-03 15:35:33 -0700460 /* Flags are always used as part of stats */
461 key->tp.flags = 0;
462
Jesse Grossccb13522011-10-25 19:26:31 -0700463 skb_reset_mac_header(skb);
464
465 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
466 * header in the linear data area.
467 */
468 eth = eth_hdr(skb);
Joe Perches8c63ff02014-02-18 11:15:45 -0800469 ether_addr_copy(key->eth.src, eth->h_source);
470 ether_addr_copy(key->eth.dst, eth->h_dest);
Jesse Grossccb13522011-10-25 19:26:31 -0700471
472 __skb_pull(skb, 2 * ETH_ALEN);
Pravin B Shelarb34df5e2013-06-13 11:11:44 -0700473 /* We are going to push all headers that we pull, so no need to
474 * update skb->csum here.
475 */
Jesse Grossccb13522011-10-25 19:26:31 -0700476
Jesse Gross07148122014-10-03 15:35:29 -0700477 key->eth.tci = 0;
Jiri Pirkodf8a39d2015-01-13 17:13:44 +0100478 if (skb_vlan_tag_present(skb))
Jesse Grossccb13522011-10-25 19:26:31 -0700479 key->eth.tci = htons(skb->vlan_tci);
480 else if (eth->h_proto == htons(ETH_P_8021Q))
481 if (unlikely(parse_vlan(skb, key)))
482 return -ENOMEM;
483
484 key->eth.type = parse_ethertype(skb);
485 if (unlikely(key->eth.type == htons(0)))
486 return -ENOMEM;
487
488 skb_reset_network_header(skb);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700489 skb_reset_mac_len(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700490 __skb_push(skb, skb->data - skb_mac_header(skb));
491
492 /* Network layer. */
493 if (key->eth.type == htons(ETH_P_IP)) {
494 struct iphdr *nh;
495 __be16 offset;
496
Jesse Grossccb13522011-10-25 19:26:31 -0700497 error = check_iphdr(skb);
498 if (unlikely(error)) {
Jesse Gross07148122014-10-03 15:35:29 -0700499 memset(&key->ip, 0, sizeof(key->ip));
500 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700501 if (error == -EINVAL) {
502 skb->transport_header = skb->network_header;
503 error = 0;
504 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700505 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700506 }
507
508 nh = ip_hdr(skb);
509 key->ipv4.addr.src = nh->saddr;
510 key->ipv4.addr.dst = nh->daddr;
511
512 key->ip.proto = nh->protocol;
513 key->ip.tos = nh->tos;
514 key->ip.ttl = nh->ttl;
515
516 offset = nh->frag_off & htons(IP_OFFSET);
517 if (offset) {
518 key->ip.frag = OVS_FRAG_TYPE_LATER;
Andy Zhou03f0d912013-08-07 20:01:00 -0700519 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700520 }
521 if (nh->frag_off & htons(IP_MF) ||
Jesse Gross07148122014-10-03 15:35:29 -0700522 skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
Jesse Grossccb13522011-10-25 19:26:31 -0700523 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Jesse Gross07148122014-10-03 15:35:29 -0700524 else
525 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700526
527 /* Transport layer. */
528 if (key->ip.proto == IPPROTO_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700529 if (tcphdr_ok(skb)) {
530 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700531 key->tp.src = tcp->source;
532 key->tp.dst = tcp->dest;
533 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700534 } else {
535 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700536 }
Jesse Gross07148122014-10-03 15:35:29 -0700537
Jesse Grossccb13522011-10-25 19:26:31 -0700538 } else if (key->ip.proto == IPPROTO_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700539 if (udphdr_ok(skb)) {
540 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700541 key->tp.src = udp->source;
542 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700543 } else {
544 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700545 }
Joe Stringera175a722013-08-22 12:30:48 -0700546 } else if (key->ip.proto == IPPROTO_SCTP) {
547 if (sctphdr_ok(skb)) {
548 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700549 key->tp.src = sctp->source;
550 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700551 } else {
552 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700553 }
Jesse Grossccb13522011-10-25 19:26:31 -0700554 } else if (key->ip.proto == IPPROTO_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700555 if (icmphdr_ok(skb)) {
556 struct icmphdr *icmp = icmp_hdr(skb);
557 /* The ICMP type and code fields use the 16-bit
558 * transport port fields, so we need to store
559 * them in 16-bit network byte order. */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700560 key->tp.src = htons(icmp->type);
561 key->tp.dst = htons(icmp->code);
Jesse Gross07148122014-10-03 15:35:29 -0700562 } else {
563 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700564 }
565 }
566
Jesse Gross07148122014-10-03 15:35:29 -0700567 } else if (key->eth.type == htons(ETH_P_ARP) ||
568 key->eth.type == htons(ETH_P_RARP)) {
Jesse Grossccb13522011-10-25 19:26:31 -0700569 struct arp_eth_header *arp;
Li RongQing389f4892014-10-17 14:03:08 +0800570 bool arp_available = arphdr_ok(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700571
572 arp = (struct arp_eth_header *)skb_network_header(skb);
573
Li RongQing389f4892014-10-17 14:03:08 +0800574 if (arp_available &&
Jesse Gross07148122014-10-03 15:35:29 -0700575 arp->ar_hrd == htons(ARPHRD_ETHER) &&
576 arp->ar_pro == htons(ETH_P_IP) &&
577 arp->ar_hln == ETH_ALEN &&
578 arp->ar_pln == 4) {
Jesse Grossccb13522011-10-25 19:26:31 -0700579
580 /* We only match on the lower 8 bits of the opcode. */
581 if (ntohs(arp->ar_op) <= 0xff)
582 key->ip.proto = ntohs(arp->ar_op);
Jesse Gross07148122014-10-03 15:35:29 -0700583 else
584 key->ip.proto = 0;
585
Mehak Mahajand04d3822012-10-30 15:50:28 -0700586 memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
587 memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
Joe Perches8c63ff02014-02-18 11:15:45 -0800588 ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
589 ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
Jesse Gross07148122014-10-03 15:35:29 -0700590 } else {
591 memset(&key->ip, 0, sizeof(key->ip));
592 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700593 }
Simon Horman25cd9ba2014-10-06 05:05:13 -0700594 } else if (eth_p_mpls(key->eth.type)) {
595 size_t stack_len = MPLS_HLEN;
596
597 /* In the presence of an MPLS label stack the end of the L2
598 * header and the beginning of the L3 header differ.
599 *
600 * Advance network_header to the beginning of the L3
601 * header. mac_len corresponds to the end of the L2 header.
602 */
603 while (1) {
604 __be32 lse;
605
606 error = check_header(skb, skb->mac_len + stack_len);
607 if (unlikely(error))
608 return 0;
609
610 memcpy(&lse, skb_network_header(skb), MPLS_HLEN);
611
612 if (stack_len == MPLS_HLEN)
613 memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN);
614
615 skb_set_network_header(skb, skb->mac_len + stack_len);
616 if (lse & htonl(MPLS_LS_S_MASK))
617 break;
618
619 stack_len += MPLS_HLEN;
620 }
Jesse Grossccb13522011-10-25 19:26:31 -0700621 } else if (key->eth.type == htons(ETH_P_IPV6)) {
622 int nh_len; /* IPv6 Header + Extensions */
623
Andy Zhou03f0d912013-08-07 20:01:00 -0700624 nh_len = parse_ipv6hdr(skb, key);
Jesse Grossccb13522011-10-25 19:26:31 -0700625 if (unlikely(nh_len < 0)) {
Jesse Gross07148122014-10-03 15:35:29 -0700626 memset(&key->ip, 0, sizeof(key->ip));
627 memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr));
Andy Zhou03f0d912013-08-07 20:01:00 -0700628 if (nh_len == -EINVAL) {
Jesse Grossccb13522011-10-25 19:26:31 -0700629 skb->transport_header = skb->network_header;
Andy Zhou03f0d912013-08-07 20:01:00 -0700630 error = 0;
631 } else {
Jesse Grossccb13522011-10-25 19:26:31 -0700632 error = nh_len;
Andy Zhou03f0d912013-08-07 20:01:00 -0700633 }
634 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700635 }
636
637 if (key->ip.frag == OVS_FRAG_TYPE_LATER)
Andy Zhou03f0d912013-08-07 20:01:00 -0700638 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700639 if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
640 key->ip.frag = OVS_FRAG_TYPE_FIRST;
641
642 /* Transport layer. */
643 if (key->ip.proto == NEXTHDR_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700644 if (tcphdr_ok(skb)) {
645 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700646 key->tp.src = tcp->source;
647 key->tp.dst = tcp->dest;
648 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700649 } else {
650 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700651 }
652 } else if (key->ip.proto == NEXTHDR_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700653 if (udphdr_ok(skb)) {
654 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700655 key->tp.src = udp->source;
656 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700657 } else {
658 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700659 }
Joe Stringera175a722013-08-22 12:30:48 -0700660 } else if (key->ip.proto == NEXTHDR_SCTP) {
661 if (sctphdr_ok(skb)) {
662 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700663 key->tp.src = sctp->source;
664 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700665 } else {
666 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700667 }
Jesse Grossccb13522011-10-25 19:26:31 -0700668 } else if (key->ip.proto == NEXTHDR_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700669 if (icmp6hdr_ok(skb)) {
Andy Zhou03f0d912013-08-07 20:01:00 -0700670 error = parse_icmpv6(skb, key, nh_len);
671 if (error)
672 return error;
Jesse Gross07148122014-10-03 15:35:29 -0700673 } else {
674 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700675 }
676 }
677 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700678 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700679}
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700680
Andy Zhou971427f32014-09-15 19:37:25 -0700681int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key)
682{
683 return key_extract(skb, key);
684}
685
Thomas Graf1d8fff92015-07-21 10:43:54 +0200686int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
Pravin B Shelar8c8b1b82014-09-15 19:28:44 -0700687 struct sk_buff *skb, struct sw_flow_key *key)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700688{
689 /* Extract metadata from packet. */
Jesse Grossf5796682014-10-03 15:35:33 -0700690 if (tun_info) {
Thomas Graf1d8fff92015-07-21 10:43:54 +0200691 memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700692
693 if (tun_info->options) {
694 BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
695 8)) - 1
696 > sizeof(key->tun_opts));
Thomas Grafd91641d2015-01-15 03:53:57 +0100697 memcpy(TUN_METADATA_OPTS(key, tun_info->options_len),
Jesse Grossf5796682014-10-03 15:35:33 -0700698 tun_info->options, tun_info->options_len);
699 key->tun_opts_len = tun_info->options_len;
700 } else {
701 key->tun_opts_len = 0;
702 }
703 } else {
704 key->tun_opts_len = 0;
Jesse Gross07148122014-10-03 15:35:29 -0700705 memset(&key->tun_key, 0, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700706 }
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700707
708 key->phy.priority = skb->priority;
709 key->phy.in_port = OVS_CB(skb)->input_vport->port_no;
710 key->phy.skb_mark = skb->mark;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700711 ovs_ct_fill_key(skb, key);
Jesse Gross07148122014-10-03 15:35:29 -0700712 key->ovs_flow_hash = 0;
713 key->recirc_id = 0;
714
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700715 return key_extract(skb, key);
716}
717
Joe Stringerc2ac6672015-08-26 11:31:52 -0700718int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700719 struct sk_buff *skb,
Jarno Rajahalme05da5892014-11-06 07:03:05 -0800720 struct sw_flow_key *key, bool log)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700721{
722 int err;
723
Pravin B Shelarb35725a2015-02-10 13:35:16 -0800724 memset(key, 0, OVS_SW_FLOW_KEY_METADATA_SIZE);
725
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700726 /* Extract metadata from netlink attributes. */
Joe Stringerc2ac6672015-08-26 11:31:52 -0700727 err = ovs_nla_get_flow_metadata(net, attr, key, log);
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700728 if (err)
729 return err;
730
731 return key_extract(skb, key);
732}