Commit | Line | Data |
---|---|---|
ccb1352e | 1 | /* |
971427f3 | 2 | * Copyright (c) 2007-2014 Nicira, Inc. |
ccb1352e JG |
3 | * |
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 | ||
ccb1352e JG |
19 | #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> | |
ccb1352e JG |
33 | #include <linux/ip.h> |
34 | #include <linux/ipv6.h> | |
a175a723 | 35 | #include <linux/sctp.h> |
e298e505 | 36 | #include <linux/smp.h> |
ccb1352e JG |
37 | #include <linux/tcp.h> |
38 | #include <linux/udp.h> | |
39 | #include <linux/icmp.h> | |
40 | #include <linux/icmpv6.h> | |
41 | #include <linux/rculist.h> | |
42 | #include <net/ip.h> | |
7d5437c7 | 43 | #include <net/ip_tunnels.h> |
ccb1352e JG |
44 | #include <net/ipv6.h> |
45 | #include <net/ndisc.h> | |
46 | ||
83c8df26 PS |
47 | #include "datapath.h" |
48 | #include "flow.h" | |
49 | #include "flow_netlink.h" | |
50 | ||
e6445719 | 51 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
03f0d916 | 52 | { |
e6445719 PS |
53 | struct timespec cur_ts; |
54 | u64 cur_ms, idle_ms; | |
03f0d916 | 55 | |
e6445719 PS |
56 | ktime_get_ts(&cur_ts); |
57 | idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); | |
58 | cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + | |
59 | cur_ts.tv_nsec / NSEC_PER_MSEC; | |
03f0d916 | 60 | |
e6445719 | 61 | return cur_ms - idle_ms; |
5828cd9a AZ |
62 | } |
63 | ||
df23e9f6 | 64 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
03f0d916 | 65 | |
ad552007 BP |
66 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
67 | struct sk_buff *skb) | |
03f0d916 | 68 | { |
e298e505 | 69 | struct flow_stats *stats; |
63e7959c | 70 | int node = numa_node_id(); |
03f0d916 | 71 | |
63e7959c | 72 | stats = rcu_dereference(flow->stats[node]); |
e298e505 | 73 | |
63e7959c JR |
74 | /* Check if already have node-specific stats. */ |
75 | if (likely(stats)) { | |
76 | spin_lock(&stats->lock); | |
77 | /* Mark if we write on the pre-allocated stats. */ | |
78 | if (node == 0 && unlikely(flow->stats_last_writer != node)) | |
79 | flow->stats_last_writer = node; | |
80 | } else { | |
81 | stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */ | |
82 | spin_lock(&stats->lock); | |
83 | ||
84 | /* If the current NUMA-node is the only writer on the | |
85 | * pre-allocated stats keep using them. | |
86 | */ | |
87 | if (unlikely(flow->stats_last_writer != node)) { | |
88 | /* A previous locker may have already allocated the | |
89 | * stats, so we need to check again. If node-specific | |
90 | * stats were already allocated, we update the pre- | |
91 | * allocated stats as we have already locked them. | |
92 | */ | |
93 | if (likely(flow->stats_last_writer != NUMA_NO_NODE) | |
8c6b00c8 | 94 | && likely(!rcu_access_pointer(flow->stats[node]))) { |
63e7959c JR |
95 | /* Try to allocate node-specific stats. */ |
96 | struct flow_stats *new_stats; | |
97 | ||
98 | new_stats = | |
99 | kmem_cache_alloc_node(flow_stats_cache, | |
100 | GFP_THISNODE | | |
101 | __GFP_NOMEMALLOC, | |
102 | node); | |
103 | if (likely(new_stats)) { | |
104 | new_stats->used = jiffies; | |
105 | new_stats->packet_count = 1; | |
106 | new_stats->byte_count = skb->len; | |
107 | new_stats->tcp_flags = tcp_flags; | |
108 | spin_lock_init(&new_stats->lock); | |
109 | ||
110 | rcu_assign_pointer(flow->stats[node], | |
111 | new_stats); | |
112 | goto unlock; | |
113 | } | |
114 | } | |
115 | flow->stats_last_writer = node; | |
116 | } | |
117 | } | |
118 | ||
e298e505 PS |
119 | stats->used = jiffies; |
120 | stats->packet_count++; | |
121 | stats->byte_count += skb->len; | |
122 | stats->tcp_flags |= tcp_flags; | |
63e7959c | 123 | unlock: |
e298e505 PS |
124 | spin_unlock(&stats->lock); |
125 | } | |
126 | ||
86ec8dba JR |
127 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
128 | void ovs_flow_stats_get(const struct sw_flow *flow, | |
129 | struct ovs_flow_stats *ovs_stats, | |
e298e505 PS |
130 | unsigned long *used, __be16 *tcp_flags) |
131 | { | |
63e7959c | 132 | int node; |
e298e505 PS |
133 | |
134 | *used = 0; | |
135 | *tcp_flags = 0; | |
136 | memset(ovs_stats, 0, sizeof(*ovs_stats)); | |
137 | ||
63e7959c | 138 | for_each_node(node) { |
86ec8dba | 139 | struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[node]); |
63e7959c JR |
140 | |
141 | if (stats) { | |
142 | /* Local CPU may write on non-local stats, so we must | |
143 | * block bottom-halves here. | |
144 | */ | |
145 | spin_lock_bh(&stats->lock); | |
146 | if (!*used || time_after(stats->used, *used)) | |
147 | *used = stats->used; | |
148 | *tcp_flags |= stats->tcp_flags; | |
149 | ovs_stats->n_packets += stats->packet_count; | |
150 | ovs_stats->n_bytes += stats->byte_count; | |
151 | spin_unlock_bh(&stats->lock); | |
152 | } | |
e298e505 | 153 | } |
e298e505 PS |
154 | } |
155 | ||
86ec8dba | 156 | /* Called with ovs_mutex. */ |
e298e505 PS |
157 | void ovs_flow_stats_clear(struct sw_flow *flow) |
158 | { | |
63e7959c JR |
159 | int node; |
160 | ||
161 | for_each_node(node) { | |
86ec8dba | 162 | struct flow_stats *stats = ovsl_dereference(flow->stats[node]); |
63e7959c JR |
163 | |
164 | if (stats) { | |
165 | spin_lock_bh(&stats->lock); | |
166 | stats->used = 0; | |
167 | stats->packet_count = 0; | |
168 | stats->byte_count = 0; | |
169 | stats->tcp_flags = 0; | |
170 | spin_unlock_bh(&stats->lock); | |
171 | } | |
172 | } | |
03f0d916 AZ |
173 | } |
174 | ||
ccb1352e JG |
175 | static int check_header(struct sk_buff *skb, int len) |
176 | { | |
177 | if (unlikely(skb->len < len)) | |
178 | return -EINVAL; | |
179 | if (unlikely(!pskb_may_pull(skb, len))) | |
180 | return -ENOMEM; | |
181 | return 0; | |
182 | } | |
183 | ||
184 | static bool arphdr_ok(struct sk_buff *skb) | |
185 | { | |
186 | return pskb_may_pull(skb, skb_network_offset(skb) + | |
187 | sizeof(struct arp_eth_header)); | |
188 | } | |
189 | ||
190 | static int check_iphdr(struct sk_buff *skb) | |
191 | { | |
192 | unsigned int nh_ofs = skb_network_offset(skb); | |
193 | unsigned int ip_len; | |
194 | int err; | |
195 | ||
196 | err = check_header(skb, nh_ofs + sizeof(struct iphdr)); | |
197 | if (unlikely(err)) | |
198 | return err; | |
199 | ||
200 | ip_len = ip_hdrlen(skb); | |
201 | if (unlikely(ip_len < sizeof(struct iphdr) || | |
202 | skb->len < nh_ofs + ip_len)) | |
203 | return -EINVAL; | |
204 | ||
205 | skb_set_transport_header(skb, nh_ofs + ip_len); | |
206 | return 0; | |
207 | } | |
208 | ||
209 | static bool tcphdr_ok(struct sk_buff *skb) | |
210 | { | |
211 | int th_ofs = skb_transport_offset(skb); | |
212 | int tcp_len; | |
213 | ||
214 | if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr)))) | |
215 | return false; | |
216 | ||
217 | tcp_len = tcp_hdrlen(skb); | |
218 | if (unlikely(tcp_len < sizeof(struct tcphdr) || | |
219 | skb->len < th_ofs + tcp_len)) | |
220 | return false; | |
221 | ||
222 | return true; | |
223 | } | |
224 | ||
225 | static bool udphdr_ok(struct sk_buff *skb) | |
226 | { | |
227 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
228 | sizeof(struct udphdr)); | |
229 | } | |
230 | ||
a175a723 JS |
231 | static bool sctphdr_ok(struct sk_buff *skb) |
232 | { | |
233 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
234 | sizeof(struct sctphdr)); | |
235 | } | |
236 | ||
ccb1352e JG |
237 | static bool icmphdr_ok(struct sk_buff *skb) |
238 | { | |
239 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
240 | sizeof(struct icmphdr)); | |
241 | } | |
242 | ||
03f0d916 | 243 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e JG |
244 | { |
245 | unsigned int nh_ofs = skb_network_offset(skb); | |
246 | unsigned int nh_len; | |
247 | int payload_ofs; | |
248 | struct ipv6hdr *nh; | |
249 | uint8_t nexthdr; | |
250 | __be16 frag_off; | |
251 | int err; | |
252 | ||
ccb1352e JG |
253 | err = check_header(skb, nh_ofs + sizeof(*nh)); |
254 | if (unlikely(err)) | |
255 | return err; | |
256 | ||
257 | nh = ipv6_hdr(skb); | |
258 | nexthdr = nh->nexthdr; | |
259 | payload_ofs = (u8 *)(nh + 1) - skb->data; | |
260 | ||
261 | key->ip.proto = NEXTHDR_NONE; | |
262 | key->ip.tos = ipv6_get_dsfield(nh); | |
263 | key->ip.ttl = nh->hop_limit; | |
264 | key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); | |
265 | key->ipv6.addr.src = nh->saddr; | |
266 | key->ipv6.addr.dst = nh->daddr; | |
267 | ||
268 | payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); | |
269 | if (unlikely(payload_ofs < 0)) | |
270 | return -EINVAL; | |
271 | ||
272 | if (frag_off) { | |
273 | if (frag_off & htons(~0x7)) | |
274 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
275 | else | |
276 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
25ef1328 PS |
277 | } else { |
278 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
279 | } |
280 | ||
281 | nh_len = payload_ofs - nh_ofs; | |
282 | skb_set_transport_header(skb, nh_ofs + nh_len); | |
283 | key->ip.proto = nexthdr; | |
284 | return nh_len; | |
285 | } | |
286 | ||
287 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
288 | { | |
289 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
290 | sizeof(struct icmp6hdr)); | |
291 | } | |
292 | ||
ccb1352e JG |
293 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
294 | { | |
295 | struct qtag_prefix { | |
296 | __be16 eth_type; /* ETH_P_8021Q */ | |
297 | __be16 tci; | |
298 | }; | |
299 | struct qtag_prefix *qp; | |
300 | ||
301 | if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16))) | |
302 | return 0; | |
303 | ||
304 | if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) + | |
305 | sizeof(__be16)))) | |
306 | return -ENOMEM; | |
307 | ||
308 | qp = (struct qtag_prefix *) skb->data; | |
309 | key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT); | |
310 | __skb_pull(skb, sizeof(struct qtag_prefix)); | |
311 | ||
312 | return 0; | |
313 | } | |
314 | ||
315 | static __be16 parse_ethertype(struct sk_buff *skb) | |
316 | { | |
317 | struct llc_snap_hdr { | |
318 | u8 dsap; /* Always 0xAA */ | |
319 | u8 ssap; /* Always 0xAA */ | |
320 | u8 ctrl; | |
321 | u8 oui[3]; | |
322 | __be16 ethertype; | |
323 | }; | |
324 | struct llc_snap_hdr *llc; | |
325 | __be16 proto; | |
326 | ||
327 | proto = *(__be16 *) skb->data; | |
328 | __skb_pull(skb, sizeof(__be16)); | |
329 | ||
e5c5d22e | 330 | if (ntohs(proto) >= ETH_P_802_3_MIN) |
ccb1352e JG |
331 | return proto; |
332 | ||
333 | if (skb->len < sizeof(struct llc_snap_hdr)) | |
334 | return htons(ETH_P_802_2); | |
335 | ||
336 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) | |
337 | return htons(0); | |
338 | ||
339 | llc = (struct llc_snap_hdr *) skb->data; | |
340 | if (llc->dsap != LLC_SAP_SNAP || | |
341 | llc->ssap != LLC_SAP_SNAP || | |
342 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
343 | return htons(ETH_P_802_2); | |
344 | ||
345 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
17b682a0 | 346 | |
e5c5d22e | 347 | if (ntohs(llc->ethertype) >= ETH_P_802_3_MIN) |
17b682a0 RL |
348 | return llc->ethertype; |
349 | ||
350 | return htons(ETH_P_802_2); | |
ccb1352e JG |
351 | } |
352 | ||
353 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, | |
03f0d916 | 354 | int nh_len) |
ccb1352e JG |
355 | { |
356 | struct icmp6hdr *icmp = icmp6_hdr(skb); | |
ccb1352e JG |
357 | |
358 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
359 | * fields, so we need to store them in 16-bit network byte order. | |
360 | */ | |
1139e241 JR |
361 | key->tp.src = htons(icmp->icmp6_type); |
362 | key->tp.dst = htons(icmp->icmp6_code); | |
25ef1328 | 363 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
ccb1352e JG |
364 | |
365 | if (icmp->icmp6_code == 0 && | |
366 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
367 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
368 | int icmp_len = skb->len - skb_transport_offset(skb); | |
369 | struct nd_msg *nd; | |
370 | int offset; | |
371 | ||
ccb1352e JG |
372 | /* In order to process neighbor discovery options, we need the |
373 | * entire packet. | |
374 | */ | |
375 | if (unlikely(icmp_len < sizeof(*nd))) | |
03f0d916 AZ |
376 | return 0; |
377 | ||
378 | if (unlikely(skb_linearize(skb))) | |
379 | return -ENOMEM; | |
ccb1352e JG |
380 | |
381 | nd = (struct nd_msg *)skb_transport_header(skb); | |
382 | key->ipv6.nd.target = nd->target; | |
ccb1352e JG |
383 | |
384 | icmp_len -= sizeof(*nd); | |
385 | offset = 0; | |
386 | while (icmp_len >= 8) { | |
387 | struct nd_opt_hdr *nd_opt = | |
388 | (struct nd_opt_hdr *)(nd->opt + offset); | |
389 | int opt_len = nd_opt->nd_opt_len * 8; | |
390 | ||
391 | if (unlikely(!opt_len || opt_len > icmp_len)) | |
03f0d916 | 392 | return 0; |
ccb1352e JG |
393 | |
394 | /* Store the link layer address if the appropriate | |
395 | * option is provided. It is considered an error if | |
396 | * the same link layer option is specified twice. | |
397 | */ | |
398 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR | |
399 | && opt_len == 8) { | |
400 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) | |
401 | goto invalid; | |
8c63ff09 JP |
402 | ether_addr_copy(key->ipv6.nd.sll, |
403 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
404 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
405 | && opt_len == 8) { | |
406 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) | |
407 | goto invalid; | |
8c63ff09 JP |
408 | ether_addr_copy(key->ipv6.nd.tll, |
409 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
410 | } |
411 | ||
412 | icmp_len -= opt_len; | |
413 | offset += opt_len; | |
414 | } | |
415 | } | |
416 | ||
03f0d916 | 417 | return 0; |
ccb1352e JG |
418 | |
419 | invalid: | |
420 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); | |
421 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
422 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
423 | ||
03f0d916 | 424 | return 0; |
ccb1352e JG |
425 | } |
426 | ||
427 | /** | |
83c8df26 | 428 | * key_extract - extracts a flow key from an Ethernet frame. |
ccb1352e JG |
429 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
430 | * Ethernet header | |
ccb1352e | 431 | * @key: output flow key |
ccb1352e JG |
432 | * |
433 | * The caller must ensure that skb->len >= ETH_HLEN. | |
434 | * | |
435 | * Returns 0 if successful, otherwise a negative errno value. | |
436 | * | |
437 | * Initializes @skb header pointers as follows: | |
438 | * | |
439 | * - skb->mac_header: the Ethernet header. | |
440 | * | |
441 | * - skb->network_header: just past the Ethernet header, or just past the | |
442 | * VLAN header, to the first byte of the Ethernet payload. | |
443 | * | |
34d94f21 | 444 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
ccb1352e JG |
445 | * on output, then just past the IP header, if one is present and |
446 | * of a correct length, otherwise the same as skb->network_header. | |
34d94f21 | 447 | * For other key->eth.type values it is left untouched. |
ccb1352e | 448 | */ |
83c8df26 | 449 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 450 | { |
03f0d916 | 451 | int error; |
ccb1352e JG |
452 | struct ethhdr *eth; |
453 | ||
f5796684 JG |
454 | /* Flags are always used as part of stats */ |
455 | key->tp.flags = 0; | |
456 | ||
ccb1352e JG |
457 | skb_reset_mac_header(skb); |
458 | ||
459 | /* Link layer. We are guaranteed to have at least the 14 byte Ethernet | |
460 | * header in the linear data area. | |
461 | */ | |
462 | eth = eth_hdr(skb); | |
8c63ff09 JP |
463 | ether_addr_copy(key->eth.src, eth->h_source); |
464 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
ccb1352e JG |
465 | |
466 | __skb_pull(skb, 2 * ETH_ALEN); | |
b34df5e8 PS |
467 | /* We are going to push all headers that we pull, so no need to |
468 | * update skb->csum here. | |
469 | */ | |
ccb1352e | 470 | |
07148121 | 471 | key->eth.tci = 0; |
ccb1352e JG |
472 | if (vlan_tx_tag_present(skb)) |
473 | key->eth.tci = htons(skb->vlan_tci); | |
474 | else if (eth->h_proto == htons(ETH_P_8021Q)) | |
475 | if (unlikely(parse_vlan(skb, key))) | |
476 | return -ENOMEM; | |
477 | ||
478 | key->eth.type = parse_ethertype(skb); | |
479 | if (unlikely(key->eth.type == htons(0))) | |
480 | return -ENOMEM; | |
481 | ||
482 | skb_reset_network_header(skb); | |
483 | __skb_push(skb, skb->data - skb_mac_header(skb)); | |
484 | ||
485 | /* Network layer. */ | |
486 | if (key->eth.type == htons(ETH_P_IP)) { | |
487 | struct iphdr *nh; | |
488 | __be16 offset; | |
489 | ||
ccb1352e JG |
490 | error = check_iphdr(skb); |
491 | if (unlikely(error)) { | |
07148121 JG |
492 | memset(&key->ip, 0, sizeof(key->ip)); |
493 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
494 | if (error == -EINVAL) { |
495 | skb->transport_header = skb->network_header; | |
496 | error = 0; | |
497 | } | |
03f0d916 | 498 | return error; |
ccb1352e JG |
499 | } |
500 | ||
501 | nh = ip_hdr(skb); | |
502 | key->ipv4.addr.src = nh->saddr; | |
503 | key->ipv4.addr.dst = nh->daddr; | |
504 | ||
505 | key->ip.proto = nh->protocol; | |
506 | key->ip.tos = nh->tos; | |
507 | key->ip.ttl = nh->ttl; | |
508 | ||
509 | offset = nh->frag_off & htons(IP_OFFSET); | |
510 | if (offset) { | |
511 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
03f0d916 | 512 | return 0; |
ccb1352e JG |
513 | } |
514 | if (nh->frag_off & htons(IP_MF) || | |
07148121 | 515 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
ccb1352e | 516 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
07148121 JG |
517 | else |
518 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
519 | |
520 | /* Transport layer. */ | |
521 | if (key->ip.proto == IPPROTO_TCP) { | |
ccb1352e JG |
522 | if (tcphdr_ok(skb)) { |
523 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
524 | key->tp.src = tcp->source; |
525 | key->tp.dst = tcp->dest; | |
526 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
527 | } else { |
528 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 529 | } |
07148121 | 530 | |
ccb1352e | 531 | } else if (key->ip.proto == IPPROTO_UDP) { |
ccb1352e JG |
532 | if (udphdr_ok(skb)) { |
533 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
534 | key->tp.src = udp->source; |
535 | key->tp.dst = udp->dest; | |
07148121 JG |
536 | } else { |
537 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 538 | } |
a175a723 JS |
539 | } else if (key->ip.proto == IPPROTO_SCTP) { |
540 | if (sctphdr_ok(skb)) { | |
541 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
542 | key->tp.src = sctp->source; |
543 | key->tp.dst = sctp->dest; | |
07148121 JG |
544 | } else { |
545 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 546 | } |
ccb1352e | 547 | } else if (key->ip.proto == IPPROTO_ICMP) { |
ccb1352e JG |
548 | if (icmphdr_ok(skb)) { |
549 | struct icmphdr *icmp = icmp_hdr(skb); | |
550 | /* The ICMP type and code fields use the 16-bit | |
551 | * transport port fields, so we need to store | |
552 | * them in 16-bit network byte order. */ | |
1139e241 JR |
553 | key->tp.src = htons(icmp->type); |
554 | key->tp.dst = htons(icmp->code); | |
07148121 JG |
555 | } else { |
556 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
557 | } |
558 | } | |
559 | ||
07148121 JG |
560 | } else if (key->eth.type == htons(ETH_P_ARP) || |
561 | key->eth.type == htons(ETH_P_RARP)) { | |
ccb1352e | 562 | struct arp_eth_header *arp; |
389f4894 | 563 | bool arp_available = arphdr_ok(skb); |
ccb1352e JG |
564 | |
565 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
566 | ||
389f4894 | 567 | if (arp_available && |
07148121 JG |
568 | arp->ar_hrd == htons(ARPHRD_ETHER) && |
569 | arp->ar_pro == htons(ETH_P_IP) && | |
570 | arp->ar_hln == ETH_ALEN && | |
571 | arp->ar_pln == 4) { | |
ccb1352e JG |
572 | |
573 | /* We only match on the lower 8 bits of the opcode. */ | |
574 | if (ntohs(arp->ar_op) <= 0xff) | |
575 | key->ip.proto = ntohs(arp->ar_op); | |
07148121 JG |
576 | else |
577 | key->ip.proto = 0; | |
578 | ||
d04d3829 MM |
579 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
580 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
8c63ff09 JP |
581 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
582 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
07148121 JG |
583 | } else { |
584 | memset(&key->ip, 0, sizeof(key->ip)); | |
585 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
586 | } |
587 | } else if (key->eth.type == htons(ETH_P_IPV6)) { | |
588 | int nh_len; /* IPv6 Header + Extensions */ | |
589 | ||
03f0d916 | 590 | nh_len = parse_ipv6hdr(skb, key); |
ccb1352e | 591 | if (unlikely(nh_len < 0)) { |
07148121 JG |
592 | memset(&key->ip, 0, sizeof(key->ip)); |
593 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
03f0d916 | 594 | if (nh_len == -EINVAL) { |
ccb1352e | 595 | skb->transport_header = skb->network_header; |
03f0d916 AZ |
596 | error = 0; |
597 | } else { | |
ccb1352e | 598 | error = nh_len; |
03f0d916 AZ |
599 | } |
600 | return error; | |
ccb1352e JG |
601 | } |
602 | ||
603 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) | |
03f0d916 | 604 | return 0; |
ccb1352e JG |
605 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
606 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
607 | ||
608 | /* Transport layer. */ | |
609 | if (key->ip.proto == NEXTHDR_TCP) { | |
ccb1352e JG |
610 | if (tcphdr_ok(skb)) { |
611 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
612 | key->tp.src = tcp->source; |
613 | key->tp.dst = tcp->dest; | |
614 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
615 | } else { |
616 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
617 | } |
618 | } else if (key->ip.proto == NEXTHDR_UDP) { | |
ccb1352e JG |
619 | if (udphdr_ok(skb)) { |
620 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
621 | key->tp.src = udp->source; |
622 | key->tp.dst = udp->dest; | |
07148121 JG |
623 | } else { |
624 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 625 | } |
a175a723 JS |
626 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
627 | if (sctphdr_ok(skb)) { | |
628 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
629 | key->tp.src = sctp->source; |
630 | key->tp.dst = sctp->dest; | |
07148121 JG |
631 | } else { |
632 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 633 | } |
ccb1352e | 634 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
ccb1352e | 635 | if (icmp6hdr_ok(skb)) { |
03f0d916 AZ |
636 | error = parse_icmpv6(skb, key, nh_len); |
637 | if (error) | |
638 | return error; | |
07148121 JG |
639 | } else { |
640 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
641 | } |
642 | } | |
643 | } | |
03f0d916 | 644 | return 0; |
ccb1352e | 645 | } |
83c8df26 | 646 | |
971427f3 AZ |
647 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
648 | { | |
649 | return key_extract(skb, key); | |
650 | } | |
651 | ||
f0b128c1 | 652 | int ovs_flow_key_extract(struct ovs_tunnel_info *tun_info, |
8c8b1b83 | 653 | struct sk_buff *skb, struct sw_flow_key *key) |
83c8df26 PS |
654 | { |
655 | /* Extract metadata from packet. */ | |
f5796684 | 656 | if (tun_info) { |
f0b128c1 | 657 | memcpy(&key->tun_key, &tun_info->tunnel, sizeof(key->tun_key)); |
f5796684 JG |
658 | |
659 | if (tun_info->options) { | |
660 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * | |
661 | 8)) - 1 | |
662 | > sizeof(key->tun_opts)); | |
663 | memcpy(GENEVE_OPTS(key, tun_info->options_len), | |
664 | tun_info->options, tun_info->options_len); | |
665 | key->tun_opts_len = tun_info->options_len; | |
666 | } else { | |
667 | key->tun_opts_len = 0; | |
668 | } | |
669 | } else { | |
670 | key->tun_opts_len = 0; | |
07148121 | 671 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
f5796684 | 672 | } |
83c8df26 PS |
673 | |
674 | key->phy.priority = skb->priority; | |
675 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
676 | key->phy.skb_mark = skb->mark; | |
07148121 JG |
677 | key->ovs_flow_hash = 0; |
678 | key->recirc_id = 0; | |
679 | ||
83c8df26 PS |
680 | return key_extract(skb, key); |
681 | } | |
682 | ||
683 | int ovs_flow_key_extract_userspace(const struct nlattr *attr, | |
684 | struct sk_buff *skb, | |
685 | struct sw_flow_key *key) | |
686 | { | |
687 | int err; | |
688 | ||
83c8df26 PS |
689 | /* Extract metadata from netlink attributes. */ |
690 | err = ovs_nla_get_flow_metadata(attr, key); | |
691 | if (err) | |
692 | return err; | |
693 | ||
694 | return key_extract(skb, key); | |
695 | } |