blob: 22087062bd1013e7c526dc95bce379bcaaebc4b9 [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>
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030032#include <linux/cpumask.h>
Jesse Grossccb13522011-10-25 19:26:31 -070033#include <linux/if_arp.h>
Jesse Grossccb13522011-10-25 19:26:31 -070034#include <linux/ip.h>
35#include <linux/ipv6.h>
Simon Horman25cd9ba2014-10-06 05:05:13 -070036#include <linux/mpls.h>
Joe Stringera175a722013-08-22 12:30:48 -070037#include <linux/sctp.h>
Pravin B Shelare298e502013-10-29 17:22:21 -070038#include <linux/smp.h>
Jesse Grossccb13522011-10-25 19:26:31 -070039#include <linux/tcp.h>
40#include <linux/udp.h>
41#include <linux/icmp.h>
42#include <linux/icmpv6.h>
43#include <linux/rculist.h>
44#include <net/ip.h>
Pravin B Shelar7d5437c2013-06-17 17:50:18 -070045#include <net/ip_tunnels.h>
Jesse Grossccb13522011-10-25 19:26:31 -070046#include <net/ipv6.h>
Simon Horman25cd9ba2014-10-06 05:05:13 -070047#include <net/mpls.h>
Jesse Grossccb13522011-10-25 19:26:31 -070048#include <net/ndisc.h>
49
Pravin B Shelara581b962015-08-29 17:44:08 -070050#include "conntrack.h"
Pravin B Shelar83c8df22014-09-15 19:20:31 -070051#include "datapath.h"
52#include "flow.h"
53#include "flow_netlink.h"
Pravin B Shelara581b962015-08-29 17:44:08 -070054#include "vport.h"
Pravin B Shelar83c8df22014-09-15 19:20:31 -070055
Pravin B Shelare6445712013-10-03 18:16:47 -070056u64 ovs_flow_used_time(unsigned long flow_jiffies)
Andy Zhou03f0d912013-08-07 20:01:00 -070057{
Pravin B Shelare6445712013-10-03 18:16:47 -070058 struct timespec cur_ts;
59 u64 cur_ms, idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070060
Pravin B Shelare6445712013-10-03 18:16:47 -070061 ktime_get_ts(&cur_ts);
62 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
63 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
64 cur_ts.tv_nsec / NSEC_PER_MSEC;
Andy Zhou03f0d912013-08-07 20:01:00 -070065
Pravin B Shelare6445712013-10-03 18:16:47 -070066 return cur_ms - idle_ms;
Andy Zhou03f0d912013-08-07 20:01:00 -070067}
68
Jarno Rajahalmedf23e9f2013-10-23 01:40:44 -070069#define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
Andy Zhou03f0d912013-08-07 20:01:00 -070070
Ben Pfaffad552002014-05-06 16:48:38 -070071void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags,
Thomas Graf12eb18f2014-11-06 06:58:52 -080072 const struct sk_buff *skb)
Andy Zhou5828cd92013-08-27 13:02:21 -070073{
Pravin B Shelare298e502013-10-29 17:22:21 -070074 struct flow_stats *stats;
Jarno Rajahalme63e79592014-03-27 12:42:54 -070075 int node = numa_node_id();
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030076 int cpu = smp_processor_id();
Jiri Pirkodf8a39d2015-01-13 17:13:44 +010077 int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Andy Zhou5828cd92013-08-27 13:02:21 -070078
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030079 stats = rcu_dereference(flow->stats[cpu]);
Pravin B Shelare298e502013-10-29 17:22:21 -070080
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030081 /* Check if already have CPU-specific stats. */
Jarno Rajahalme63e79592014-03-27 12:42:54 -070082 if (likely(stats)) {
83 spin_lock(&stats->lock);
84 /* Mark if we write on the pre-allocated stats. */
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030085 if (cpu == 0 && unlikely(flow->stats_last_writer != cpu))
86 flow->stats_last_writer = cpu;
Jarno Rajahalme63e79592014-03-27 12:42:54 -070087 } else {
88 stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */
89 spin_lock(&stats->lock);
90
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030091 /* If the current CPU is the only writer on the
Jarno Rajahalme63e79592014-03-27 12:42:54 -070092 * pre-allocated stats keep using them.
93 */
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030094 if (unlikely(flow->stats_last_writer != cpu)) {
Jarno Rajahalme63e79592014-03-27 12:42:54 -070095 /* A previous locker may have already allocated the
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -030096 * stats, so we need to check again. If CPU-specific
Jarno Rajahalme63e79592014-03-27 12:42:54 -070097 * stats were already allocated, we update the pre-
98 * allocated stats as we have already locked them.
99 */
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300100 if (likely(flow->stats_last_writer != -1) &&
101 likely(!rcu_access_pointer(flow->stats[cpu]))) {
102 /* Try to allocate CPU-specific stats. */
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700103 struct flow_stats *new_stats;
104
105 new_stats =
106 kmem_cache_alloc_node(flow_stats_cache,
David Rientjes4167e9b2015-04-14 15:46:55 -0700107 GFP_NOWAIT |
108 __GFP_THISNODE |
109 __GFP_NOWARN |
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700110 __GFP_NOMEMALLOC,
111 node);
112 if (likely(new_stats)) {
113 new_stats->used = jiffies;
114 new_stats->packet_count = 1;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800115 new_stats->byte_count = len;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700116 new_stats->tcp_flags = tcp_flags;
117 spin_lock_init(&new_stats->lock);
118
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300119 rcu_assign_pointer(flow->stats[cpu],
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700120 new_stats);
121 goto unlock;
122 }
123 }
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300124 flow->stats_last_writer = cpu;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700125 }
126 }
127
Pravin B Shelare298e502013-10-29 17:22:21 -0700128 stats->used = jiffies;
129 stats->packet_count++;
Ben Pfaff24cc59d2014-12-31 08:45:46 -0800130 stats->byte_count += len;
Pravin B Shelare298e502013-10-29 17:22:21 -0700131 stats->tcp_flags |= tcp_flags;
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700132unlock:
Pravin B Shelare298e502013-10-29 17:22:21 -0700133 spin_unlock(&stats->lock);
134}
135
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700136/* Must be called with rcu_read_lock or ovs_mutex. */
137void ovs_flow_stats_get(const struct sw_flow *flow,
138 struct ovs_flow_stats *ovs_stats,
Pravin B Shelare298e502013-10-29 17:22:21 -0700139 unsigned long *used, __be16 *tcp_flags)
140{
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300141 int cpu;
Pravin B Shelare298e502013-10-29 17:22:21 -0700142
143 *used = 0;
144 *tcp_flags = 0;
145 memset(ovs_stats, 0, sizeof(*ovs_stats));
146
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300147 /* We open code this to make sure cpu 0 is always considered */
148 for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) {
149 struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[cpu]);
Pravin B Shelare298e502013-10-29 17:22:21 -0700150
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700151 if (stats) {
152 /* Local CPU may write on non-local stats, so we must
153 * block bottom-halves here.
154 */
155 spin_lock_bh(&stats->lock);
156 if (!*used || time_after(stats->used, *used))
157 *used = stats->used;
158 *tcp_flags |= stats->tcp_flags;
159 ovs_stats->n_packets += stats->packet_count;
160 ovs_stats->n_bytes += stats->byte_count;
161 spin_unlock_bh(&stats->lock);
162 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700163 }
Pravin B Shelare298e502013-10-29 17:22:21 -0700164}
165
Jarno Rajahalme86ec8db2014-05-05 14:17:28 -0700166/* Called with ovs_mutex. */
Pravin B Shelare298e502013-10-29 17:22:21 -0700167void ovs_flow_stats_clear(struct sw_flow *flow)
168{
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300169 int cpu;
Pravin B Shelare298e502013-10-29 17:22:21 -0700170
Thadeu Lima de Souza Cascardodb74a332016-09-15 19:11:53 -0300171 /* We open code this to make sure cpu 0 is always considered */
172 for (cpu = 0; cpu < nr_cpu_ids; cpu = cpumask_next(cpu, cpu_possible_mask)) {
173 struct flow_stats *stats = ovsl_dereference(flow->stats[cpu]);
Jarno Rajahalme23dabf82014-03-27 12:35:23 -0700174
Jarno Rajahalme63e79592014-03-27 12:42:54 -0700175 if (stats) {
176 spin_lock_bh(&stats->lock);
177 stats->used = 0;
178 stats->packet_count = 0;
179 stats->byte_count = 0;
180 stats->tcp_flags = 0;
181 spin_unlock_bh(&stats->lock);
182 }
183 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700184}
185
Jesse Grossccb13522011-10-25 19:26:31 -0700186static int check_header(struct sk_buff *skb, int len)
187{
188 if (unlikely(skb->len < len))
189 return -EINVAL;
190 if (unlikely(!pskb_may_pull(skb, len)))
191 return -ENOMEM;
192 return 0;
193}
194
195static bool arphdr_ok(struct sk_buff *skb)
196{
197 return pskb_may_pull(skb, skb_network_offset(skb) +
198 sizeof(struct arp_eth_header));
199}
200
201static int check_iphdr(struct sk_buff *skb)
202{
203 unsigned int nh_ofs = skb_network_offset(skb);
204 unsigned int ip_len;
205 int err;
206
207 err = check_header(skb, nh_ofs + sizeof(struct iphdr));
208 if (unlikely(err))
209 return err;
210
211 ip_len = ip_hdrlen(skb);
212 if (unlikely(ip_len < sizeof(struct iphdr) ||
213 skb->len < nh_ofs + ip_len))
214 return -EINVAL;
215
216 skb_set_transport_header(skb, nh_ofs + ip_len);
217 return 0;
218}
219
220static bool tcphdr_ok(struct sk_buff *skb)
221{
222 int th_ofs = skb_transport_offset(skb);
223 int tcp_len;
224
225 if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr))))
226 return false;
227
228 tcp_len = tcp_hdrlen(skb);
229 if (unlikely(tcp_len < sizeof(struct tcphdr) ||
230 skb->len < th_ofs + tcp_len))
231 return false;
232
233 return true;
234}
235
236static bool udphdr_ok(struct sk_buff *skb)
237{
238 return pskb_may_pull(skb, skb_transport_offset(skb) +
239 sizeof(struct udphdr));
240}
241
Joe Stringera175a722013-08-22 12:30:48 -0700242static bool sctphdr_ok(struct sk_buff *skb)
243{
244 return pskb_may_pull(skb, skb_transport_offset(skb) +
245 sizeof(struct sctphdr));
246}
247
Jesse Grossccb13522011-10-25 19:26:31 -0700248static bool icmphdr_ok(struct sk_buff *skb)
249{
250 return pskb_may_pull(skb, skb_transport_offset(skb) +
251 sizeof(struct icmphdr));
252}
253
Andy Zhou03f0d912013-08-07 20:01:00 -0700254static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700255{
256 unsigned int nh_ofs = skb_network_offset(skb);
257 unsigned int nh_len;
258 int payload_ofs;
259 struct ipv6hdr *nh;
260 uint8_t nexthdr;
261 __be16 frag_off;
262 int err;
263
Jesse Grossccb13522011-10-25 19:26:31 -0700264 err = check_header(skb, nh_ofs + sizeof(*nh));
265 if (unlikely(err))
266 return err;
267
268 nh = ipv6_hdr(skb);
269 nexthdr = nh->nexthdr;
270 payload_ofs = (u8 *)(nh + 1) - skb->data;
271
272 key->ip.proto = NEXTHDR_NONE;
273 key->ip.tos = ipv6_get_dsfield(nh);
274 key->ip.ttl = nh->hop_limit;
275 key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL);
276 key->ipv6.addr.src = nh->saddr;
277 key->ipv6.addr.dst = nh->daddr;
278
279 payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off);
Jesse Grossccb13522011-10-25 19:26:31 -0700280
281 if (frag_off) {
282 if (frag_off & htons(~0x7))
283 key->ip.frag = OVS_FRAG_TYPE_LATER;
284 else
285 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700286 } else {
287 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700288 }
289
Simon Hormanc30da492015-08-29 09:02:21 +0900290 /* Delayed handling of error in ipv6_skip_exthdr() as it
291 * always sets frag_off to a valid value which may be
292 * used to set key->ip.frag above.
293 */
294 if (unlikely(payload_ofs < 0))
295 return -EPROTO;
296
Jesse Grossccb13522011-10-25 19:26:31 -0700297 nh_len = payload_ofs - nh_ofs;
298 skb_set_transport_header(skb, nh_ofs + nh_len);
299 key->ip.proto = nexthdr;
300 return nh_len;
301}
302
303static bool icmp6hdr_ok(struct sk_buff *skb)
304{
305 return pskb_may_pull(skb, skb_transport_offset(skb) +
306 sizeof(struct icmp6hdr));
307}
308
Eric Garver018c1dd2016-09-07 12:56:59 -0400309/**
310 * Parse vlan tag from vlan header.
311 * Returns ERROR on memory error.
312 * Returns 0 if it encounters a non-vlan or incomplete packet.
313 * Returns 1 after successfully parsing vlan tag.
314 */
315static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh)
Jesse Grossccb13522011-10-25 19:26:31 -0700316{
Eric Garver018c1dd2016-09-07 12:56:59 -0400317 struct vlan_head *vh = (struct vlan_head *)skb->data;
Jesse Grossccb13522011-10-25 19:26:31 -0700318
Eric Garver018c1dd2016-09-07 12:56:59 -0400319 if (likely(!eth_type_vlan(vh->tpid)))
Jesse Grossccb13522011-10-25 19:26:31 -0700320 return 0;
321
Eric Garver018c1dd2016-09-07 12:56:59 -0400322 if (unlikely(skb->len < sizeof(struct vlan_head) + sizeof(__be16)))
323 return 0;
324
325 if (unlikely(!pskb_may_pull(skb, sizeof(struct vlan_head) +
326 sizeof(__be16))))
Jesse Grossccb13522011-10-25 19:26:31 -0700327 return -ENOMEM;
328
Eric Garver018c1dd2016-09-07 12:56:59 -0400329 vh = (struct vlan_head *)skb->data;
330 key_vh->tci = vh->tci | htons(VLAN_TAG_PRESENT);
331 key_vh->tpid = vh->tpid;
332
333 __skb_pull(skb, sizeof(struct vlan_head));
334 return 1;
335}
336
337static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
338{
339 int res;
340
341 key->eth.vlan.tci = 0;
342 key->eth.vlan.tpid = 0;
343 key->eth.cvlan.tci = 0;
344 key->eth.cvlan.tpid = 0;
345
Jiri Benc20ecf1e2016-10-10 17:02:42 +0200346 if (skb_vlan_tag_present(skb)) {
Eric Garver018c1dd2016-09-07 12:56:59 -0400347 key->eth.vlan.tci = htons(skb->vlan_tci);
348 key->eth.vlan.tpid = skb->vlan_proto;
349 } else {
350 /* Parse outer vlan tag in the non-accelerated case. */
351 res = parse_vlan_tag(skb, &key->eth.vlan);
352 if (res <= 0)
353 return res;
354 }
355
356 /* Parse inner vlan tag. */
357 res = parse_vlan_tag(skb, &key->eth.cvlan);
358 if (res <= 0)
359 return res;
Jesse Grossccb13522011-10-25 19:26:31 -0700360
361 return 0;
362}
363
364static __be16 parse_ethertype(struct sk_buff *skb)
365{
366 struct llc_snap_hdr {
367 u8 dsap; /* Always 0xAA */
368 u8 ssap; /* Always 0xAA */
369 u8 ctrl;
370 u8 oui[3];
371 __be16 ethertype;
372 };
373 struct llc_snap_hdr *llc;
374 __be16 proto;
375
376 proto = *(__be16 *) skb->data;
377 __skb_pull(skb, sizeof(__be16));
378
Alexander Duyck6713fc92015-05-04 14:34:05 -0700379 if (eth_proto_is_802_3(proto))
Jesse Grossccb13522011-10-25 19:26:31 -0700380 return proto;
381
382 if (skb->len < sizeof(struct llc_snap_hdr))
383 return htons(ETH_P_802_2);
384
385 if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr))))
386 return htons(0);
387
388 llc = (struct llc_snap_hdr *) skb->data;
389 if (llc->dsap != LLC_SAP_SNAP ||
390 llc->ssap != LLC_SAP_SNAP ||
391 (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0)
392 return htons(ETH_P_802_2);
393
394 __skb_pull(skb, sizeof(struct llc_snap_hdr));
Rich Lane17b682a2013-02-19 11:10:30 -0800395
Alexander Duyck6713fc92015-05-04 14:34:05 -0700396 if (eth_proto_is_802_3(llc->ethertype))
Rich Lane17b682a2013-02-19 11:10:30 -0800397 return llc->ethertype;
398
399 return htons(ETH_P_802_2);
Jesse Grossccb13522011-10-25 19:26:31 -0700400}
401
402static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key,
Andy Zhou03f0d912013-08-07 20:01:00 -0700403 int nh_len)
Jesse Grossccb13522011-10-25 19:26:31 -0700404{
405 struct icmp6hdr *icmp = icmp6_hdr(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700406
407 /* The ICMPv6 type and code fields use the 16-bit transport port
408 * fields, so we need to store them in 16-bit network byte order.
409 */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700410 key->tp.src = htons(icmp->icmp6_type);
411 key->tp.dst = htons(icmp->icmp6_code);
Pravin B Shelar25ef1322014-10-17 13:56:31 -0700412 memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd));
Jesse Grossccb13522011-10-25 19:26:31 -0700413
414 if (icmp->icmp6_code == 0 &&
415 (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION ||
416 icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) {
417 int icmp_len = skb->len - skb_transport_offset(skb);
418 struct nd_msg *nd;
419 int offset;
420
Jesse Grossccb13522011-10-25 19:26:31 -0700421 /* In order to process neighbor discovery options, we need the
422 * entire packet.
423 */
424 if (unlikely(icmp_len < sizeof(*nd)))
Andy Zhou03f0d912013-08-07 20:01:00 -0700425 return 0;
426
427 if (unlikely(skb_linearize(skb)))
428 return -ENOMEM;
Jesse Grossccb13522011-10-25 19:26:31 -0700429
430 nd = (struct nd_msg *)skb_transport_header(skb);
431 key->ipv6.nd.target = nd->target;
Jesse Grossccb13522011-10-25 19:26:31 -0700432
433 icmp_len -= sizeof(*nd);
434 offset = 0;
435 while (icmp_len >= 8) {
436 struct nd_opt_hdr *nd_opt =
437 (struct nd_opt_hdr *)(nd->opt + offset);
438 int opt_len = nd_opt->nd_opt_len * 8;
439
440 if (unlikely(!opt_len || opt_len > icmp_len))
Andy Zhou03f0d912013-08-07 20:01:00 -0700441 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700442
443 /* Store the link layer address if the appropriate
444 * option is provided. It is considered an error if
445 * the same link layer option is specified twice.
446 */
447 if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR
448 && opt_len == 8) {
449 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll)))
450 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800451 ether_addr_copy(key->ipv6.nd.sll,
452 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700453 } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR
454 && opt_len == 8) {
455 if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll)))
456 goto invalid;
Joe Perches8c63ff02014-02-18 11:15:45 -0800457 ether_addr_copy(key->ipv6.nd.tll,
458 &nd->opt[offset+sizeof(*nd_opt)]);
Jesse Grossccb13522011-10-25 19:26:31 -0700459 }
460
461 icmp_len -= opt_len;
462 offset += opt_len;
463 }
464 }
465
Andy Zhou03f0d912013-08-07 20:01:00 -0700466 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700467
468invalid:
469 memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target));
470 memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll));
471 memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll));
472
Andy Zhou03f0d912013-08-07 20:01:00 -0700473 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700474}
475
476/**
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700477 * key_extract - extracts a flow key from an Ethernet frame.
Jesse Grossccb13522011-10-25 19:26:31 -0700478 * @skb: sk_buff that contains the frame, with skb->data pointing to the
479 * Ethernet header
Jesse Grossccb13522011-10-25 19:26:31 -0700480 * @key: output flow key
Jesse Grossccb13522011-10-25 19:26:31 -0700481 *
482 * The caller must ensure that skb->len >= ETH_HLEN.
483 *
484 * Returns 0 if successful, otherwise a negative errno value.
485 *
486 * Initializes @skb header pointers as follows:
487 *
488 * - skb->mac_header: the Ethernet header.
489 *
490 * - skb->network_header: just past the Ethernet header, or just past the
491 * VLAN header, to the first byte of the Ethernet payload.
492 *
Lorand Jakab34d94f22013-06-03 10:01:14 -0700493 * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6
Jesse Grossccb13522011-10-25 19:26:31 -0700494 * on output, then just past the IP header, if one is present and
495 * of a correct length, otherwise the same as skb->network_header.
Lorand Jakab34d94f22013-06-03 10:01:14 -0700496 * For other key->eth.type values it is left untouched.
Jesse Grossccb13522011-10-25 19:26:31 -0700497 */
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700498static int key_extract(struct sk_buff *skb, struct sw_flow_key *key)
Jesse Grossccb13522011-10-25 19:26:31 -0700499{
Andy Zhou03f0d912013-08-07 20:01:00 -0700500 int error;
Jesse Grossccb13522011-10-25 19:26:31 -0700501 struct ethhdr *eth;
502
Jesse Grossf5796682014-10-03 15:35:33 -0700503 /* Flags are always used as part of stats */
504 key->tp.flags = 0;
505
Jesse Grossccb13522011-10-25 19:26:31 -0700506 skb_reset_mac_header(skb);
507
508 /* Link layer. We are guaranteed to have at least the 14 byte Ethernet
509 * header in the linear data area.
510 */
511 eth = eth_hdr(skb);
Joe Perches8c63ff02014-02-18 11:15:45 -0800512 ether_addr_copy(key->eth.src, eth->h_source);
513 ether_addr_copy(key->eth.dst, eth->h_dest);
Jesse Grossccb13522011-10-25 19:26:31 -0700514
515 __skb_pull(skb, 2 * ETH_ALEN);
Pravin B Shelarb34df5e2013-06-13 11:11:44 -0700516 /* We are going to push all headers that we pull, so no need to
517 * update skb->csum here.
518 */
Jesse Grossccb13522011-10-25 19:26:31 -0700519
Eric Garver018c1dd2016-09-07 12:56:59 -0400520 if (unlikely(parse_vlan(skb, key)))
521 return -ENOMEM;
Jesse Grossccb13522011-10-25 19:26:31 -0700522
523 key->eth.type = parse_ethertype(skb);
524 if (unlikely(key->eth.type == htons(0)))
525 return -ENOMEM;
526
527 skb_reset_network_header(skb);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700528 skb_reset_mac_len(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700529 __skb_push(skb, skb->data - skb_mac_header(skb));
530
531 /* Network layer. */
532 if (key->eth.type == htons(ETH_P_IP)) {
533 struct iphdr *nh;
534 __be16 offset;
535
Jesse Grossccb13522011-10-25 19:26:31 -0700536 error = check_iphdr(skb);
537 if (unlikely(error)) {
Jesse Gross07148122014-10-03 15:35:29 -0700538 memset(&key->ip, 0, sizeof(key->ip));
539 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700540 if (error == -EINVAL) {
541 skb->transport_header = skb->network_header;
542 error = 0;
543 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700544 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700545 }
546
547 nh = ip_hdr(skb);
548 key->ipv4.addr.src = nh->saddr;
549 key->ipv4.addr.dst = nh->daddr;
550
551 key->ip.proto = nh->protocol;
552 key->ip.tos = nh->tos;
553 key->ip.ttl = nh->ttl;
554
555 offset = nh->frag_off & htons(IP_OFFSET);
556 if (offset) {
557 key->ip.frag = OVS_FRAG_TYPE_LATER;
Andy Zhou03f0d912013-08-07 20:01:00 -0700558 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700559 }
560 if (nh->frag_off & htons(IP_MF) ||
Jesse Gross07148122014-10-03 15:35:29 -0700561 skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
Jesse Grossccb13522011-10-25 19:26:31 -0700562 key->ip.frag = OVS_FRAG_TYPE_FIRST;
Jesse Gross07148122014-10-03 15:35:29 -0700563 else
564 key->ip.frag = OVS_FRAG_TYPE_NONE;
Jesse Grossccb13522011-10-25 19:26:31 -0700565
566 /* Transport layer. */
567 if (key->ip.proto == IPPROTO_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700568 if (tcphdr_ok(skb)) {
569 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700570 key->tp.src = tcp->source;
571 key->tp.dst = tcp->dest;
572 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700573 } else {
574 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700575 }
Jesse Gross07148122014-10-03 15:35:29 -0700576
Jesse Grossccb13522011-10-25 19:26:31 -0700577 } else if (key->ip.proto == IPPROTO_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700578 if (udphdr_ok(skb)) {
579 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700580 key->tp.src = udp->source;
581 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700582 } else {
583 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700584 }
Joe Stringera175a722013-08-22 12:30:48 -0700585 } else if (key->ip.proto == IPPROTO_SCTP) {
586 if (sctphdr_ok(skb)) {
587 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700588 key->tp.src = sctp->source;
589 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700590 } else {
591 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700592 }
Jesse Grossccb13522011-10-25 19:26:31 -0700593 } else if (key->ip.proto == IPPROTO_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700594 if (icmphdr_ok(skb)) {
595 struct icmphdr *icmp = icmp_hdr(skb);
596 /* The ICMP type and code fields use the 16-bit
597 * transport port fields, so we need to store
598 * them in 16-bit network byte order. */
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700599 key->tp.src = htons(icmp->type);
600 key->tp.dst = htons(icmp->code);
Jesse Gross07148122014-10-03 15:35:29 -0700601 } else {
602 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700603 }
604 }
605
Jesse Gross07148122014-10-03 15:35:29 -0700606 } else if (key->eth.type == htons(ETH_P_ARP) ||
607 key->eth.type == htons(ETH_P_RARP)) {
Jesse Grossccb13522011-10-25 19:26:31 -0700608 struct arp_eth_header *arp;
Li RongQing389f4892014-10-17 14:03:08 +0800609 bool arp_available = arphdr_ok(skb);
Jesse Grossccb13522011-10-25 19:26:31 -0700610
611 arp = (struct arp_eth_header *)skb_network_header(skb);
612
Li RongQing389f4892014-10-17 14:03:08 +0800613 if (arp_available &&
Jesse Gross07148122014-10-03 15:35:29 -0700614 arp->ar_hrd == htons(ARPHRD_ETHER) &&
615 arp->ar_pro == htons(ETH_P_IP) &&
616 arp->ar_hln == ETH_ALEN &&
617 arp->ar_pln == 4) {
Jesse Grossccb13522011-10-25 19:26:31 -0700618
619 /* We only match on the lower 8 bits of the opcode. */
620 if (ntohs(arp->ar_op) <= 0xff)
621 key->ip.proto = ntohs(arp->ar_op);
Jesse Gross07148122014-10-03 15:35:29 -0700622 else
623 key->ip.proto = 0;
624
Mehak Mahajand04d3822012-10-30 15:50:28 -0700625 memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src));
626 memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst));
Joe Perches8c63ff02014-02-18 11:15:45 -0800627 ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha);
628 ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha);
Jesse Gross07148122014-10-03 15:35:29 -0700629 } else {
630 memset(&key->ip, 0, sizeof(key->ip));
631 memset(&key->ipv4, 0, sizeof(key->ipv4));
Jesse Grossccb13522011-10-25 19:26:31 -0700632 }
Simon Horman25cd9ba2014-10-06 05:05:13 -0700633 } else if (eth_p_mpls(key->eth.type)) {
634 size_t stack_len = MPLS_HLEN;
635
Jiri Bencf7d49bc2016-09-30 19:08:05 +0200636 skb_set_inner_network_header(skb, skb->mac_len);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700637 while (1) {
638 __be32 lse;
639
640 error = check_header(skb, skb->mac_len + stack_len);
641 if (unlikely(error))
642 return 0;
643
Jiri Bencf7d49bc2016-09-30 19:08:05 +0200644 memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700645
646 if (stack_len == MPLS_HLEN)
647 memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN);
648
Jiri Bencf7d49bc2016-09-30 19:08:05 +0200649 skb_set_inner_network_header(skb, skb->mac_len + stack_len);
Simon Horman25cd9ba2014-10-06 05:05:13 -0700650 if (lse & htonl(MPLS_LS_S_MASK))
651 break;
652
653 stack_len += MPLS_HLEN;
654 }
Jesse Grossccb13522011-10-25 19:26:31 -0700655 } else if (key->eth.type == htons(ETH_P_IPV6)) {
656 int nh_len; /* IPv6 Header + Extensions */
657
Andy Zhou03f0d912013-08-07 20:01:00 -0700658 nh_len = parse_ipv6hdr(skb, key);
Jesse Grossccb13522011-10-25 19:26:31 -0700659 if (unlikely(nh_len < 0)) {
Simon Hormanc30da492015-08-29 09:02:21 +0900660 switch (nh_len) {
661 case -EINVAL:
662 memset(&key->ip, 0, sizeof(key->ip));
663 memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr));
664 /* fall-through */
665 case -EPROTO:
Jesse Grossccb13522011-10-25 19:26:31 -0700666 skb->transport_header = skb->network_header;
Andy Zhou03f0d912013-08-07 20:01:00 -0700667 error = 0;
Simon Hormanc30da492015-08-29 09:02:21 +0900668 break;
669 default:
Jesse Grossccb13522011-10-25 19:26:31 -0700670 error = nh_len;
Andy Zhou03f0d912013-08-07 20:01:00 -0700671 }
672 return error;
Jesse Grossccb13522011-10-25 19:26:31 -0700673 }
674
675 if (key->ip.frag == OVS_FRAG_TYPE_LATER)
Andy Zhou03f0d912013-08-07 20:01:00 -0700676 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700677 if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP)
678 key->ip.frag = OVS_FRAG_TYPE_FIRST;
679
680 /* Transport layer. */
681 if (key->ip.proto == NEXTHDR_TCP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700682 if (tcphdr_ok(skb)) {
683 struct tcphdr *tcp = tcp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700684 key->tp.src = tcp->source;
685 key->tp.dst = tcp->dest;
686 key->tp.flags = TCP_FLAGS_BE16(tcp);
Jesse Gross07148122014-10-03 15:35:29 -0700687 } else {
688 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700689 }
690 } else if (key->ip.proto == NEXTHDR_UDP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700691 if (udphdr_ok(skb)) {
692 struct udphdr *udp = udp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700693 key->tp.src = udp->source;
694 key->tp.dst = udp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700695 } else {
696 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700697 }
Joe Stringera175a722013-08-22 12:30:48 -0700698 } else if (key->ip.proto == NEXTHDR_SCTP) {
699 if (sctphdr_ok(skb)) {
700 struct sctphdr *sctp = sctp_hdr(skb);
Jarno Rajahalme1139e242014-05-05 09:54:49 -0700701 key->tp.src = sctp->source;
702 key->tp.dst = sctp->dest;
Jesse Gross07148122014-10-03 15:35:29 -0700703 } else {
704 memset(&key->tp, 0, sizeof(key->tp));
Joe Stringera175a722013-08-22 12:30:48 -0700705 }
Jesse Grossccb13522011-10-25 19:26:31 -0700706 } else if (key->ip.proto == NEXTHDR_ICMP) {
Jesse Grossccb13522011-10-25 19:26:31 -0700707 if (icmp6hdr_ok(skb)) {
Andy Zhou03f0d912013-08-07 20:01:00 -0700708 error = parse_icmpv6(skb, key, nh_len);
709 if (error)
710 return error;
Jesse Gross07148122014-10-03 15:35:29 -0700711 } else {
712 memset(&key->tp, 0, sizeof(key->tp));
Jesse Grossccb13522011-10-25 19:26:31 -0700713 }
714 }
715 }
Andy Zhou03f0d912013-08-07 20:01:00 -0700716 return 0;
Jesse Grossccb13522011-10-25 19:26:31 -0700717}
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700718
Andy Zhou971427f32014-09-15 19:37:25 -0700719int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key)
720{
721 return key_extract(skb, key);
722}
723
Thomas Graf1d8fff92015-07-21 10:43:54 +0200724int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info,
Pravin B Shelar8c8b1b82014-09-15 19:28:44 -0700725 struct sk_buff *skb, struct sw_flow_key *key)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700726{
727 /* Extract metadata from packet. */
Jesse Grossf5796682014-10-03 15:35:33 -0700728 if (tun_info) {
Jiri Benc00a93ba2015-10-05 13:09:46 +0200729 key->tun_proto = ip_tunnel_info_af(tun_info);
Thomas Graf1d8fff92015-07-21 10:43:54 +0200730 memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700731
Pravin B Shelar4c222792015-08-30 18:09:38 -0700732 if (tun_info->options_len) {
Jesse Grossf5796682014-10-03 15:35:33 -0700733 BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) *
734 8)) - 1
735 > sizeof(key->tun_opts));
Pravin B Shelar4c222792015-08-30 18:09:38 -0700736
737 ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len),
738 tun_info);
Jesse Grossf5796682014-10-03 15:35:33 -0700739 key->tun_opts_len = tun_info->options_len;
740 } else {
741 key->tun_opts_len = 0;
742 }
743 } else {
Jiri Benc00a93ba2015-10-05 13:09:46 +0200744 key->tun_proto = 0;
Jesse Grossf5796682014-10-03 15:35:33 -0700745 key->tun_opts_len = 0;
Jesse Gross07148122014-10-03 15:35:29 -0700746 memset(&key->tun_key, 0, sizeof(key->tun_key));
Jesse Grossf5796682014-10-03 15:35:33 -0700747 }
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700748
749 key->phy.priority = skb->priority;
750 key->phy.in_port = OVS_CB(skb)->input_vport->port_no;
751 key->phy.skb_mark = skb->mark;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700752 ovs_ct_fill_key(skb, key);
Jesse Gross07148122014-10-03 15:35:29 -0700753 key->ovs_flow_hash = 0;
754 key->recirc_id = 0;
755
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700756 return key_extract(skb, key);
757}
758
Joe Stringerc2ac6672015-08-26 11:31:52 -0700759int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr,
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700760 struct sk_buff *skb,
Jarno Rajahalme05da5892014-11-06 07:03:05 -0800761 struct sw_flow_key *key, bool log)
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700762{
763 int err;
764
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700765 /* Extract metadata from netlink attributes. */
Joe Stringerc2ac6672015-08-26 11:31:52 -0700766 err = ovs_nla_get_flow_metadata(net, attr, key, log);
Pravin B Shelar83c8df22014-09-15 19:20:31 -0700767 if (err)
768 return err;
769
770 return key_extract(skb, key);
771}