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; | |
277 | } | |
278 | ||
279 | nh_len = payload_ofs - nh_ofs; | |
280 | skb_set_transport_header(skb, nh_ofs + nh_len); | |
281 | key->ip.proto = nexthdr; | |
282 | return nh_len; | |
283 | } | |
284 | ||
285 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
286 | { | |
287 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
288 | sizeof(struct icmp6hdr)); | |
289 | } | |
290 | ||
ccb1352e JG |
291 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
292 | { | |
293 | struct qtag_prefix { | |
294 | __be16 eth_type; /* ETH_P_8021Q */ | |
295 | __be16 tci; | |
296 | }; | |
297 | struct qtag_prefix *qp; | |
298 | ||
299 | if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16))) | |
300 | return 0; | |
301 | ||
302 | if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) + | |
303 | sizeof(__be16)))) | |
304 | return -ENOMEM; | |
305 | ||
306 | qp = (struct qtag_prefix *) skb->data; | |
307 | key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT); | |
308 | __skb_pull(skb, sizeof(struct qtag_prefix)); | |
309 | ||
310 | return 0; | |
311 | } | |
312 | ||
313 | static __be16 parse_ethertype(struct sk_buff *skb) | |
314 | { | |
315 | struct llc_snap_hdr { | |
316 | u8 dsap; /* Always 0xAA */ | |
317 | u8 ssap; /* Always 0xAA */ | |
318 | u8 ctrl; | |
319 | u8 oui[3]; | |
320 | __be16 ethertype; | |
321 | }; | |
322 | struct llc_snap_hdr *llc; | |
323 | __be16 proto; | |
324 | ||
325 | proto = *(__be16 *) skb->data; | |
326 | __skb_pull(skb, sizeof(__be16)); | |
327 | ||
e5c5d22e | 328 | if (ntohs(proto) >= ETH_P_802_3_MIN) |
ccb1352e JG |
329 | return proto; |
330 | ||
331 | if (skb->len < sizeof(struct llc_snap_hdr)) | |
332 | return htons(ETH_P_802_2); | |
333 | ||
334 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) | |
335 | return htons(0); | |
336 | ||
337 | llc = (struct llc_snap_hdr *) skb->data; | |
338 | if (llc->dsap != LLC_SAP_SNAP || | |
339 | llc->ssap != LLC_SAP_SNAP || | |
340 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
341 | return htons(ETH_P_802_2); | |
342 | ||
343 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
17b682a0 | 344 | |
e5c5d22e | 345 | if (ntohs(llc->ethertype) >= ETH_P_802_3_MIN) |
17b682a0 RL |
346 | return llc->ethertype; |
347 | ||
348 | return htons(ETH_P_802_2); | |
ccb1352e JG |
349 | } |
350 | ||
351 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, | |
03f0d916 | 352 | int nh_len) |
ccb1352e JG |
353 | { |
354 | struct icmp6hdr *icmp = icmp6_hdr(skb); | |
ccb1352e JG |
355 | |
356 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
357 | * fields, so we need to store them in 16-bit network byte order. | |
358 | */ | |
1139e241 JR |
359 | key->tp.src = htons(icmp->icmp6_type); |
360 | key->tp.dst = htons(icmp->icmp6_code); | |
ccb1352e JG |
361 | |
362 | if (icmp->icmp6_code == 0 && | |
363 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
364 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
365 | int icmp_len = skb->len - skb_transport_offset(skb); | |
366 | struct nd_msg *nd; | |
367 | int offset; | |
368 | ||
ccb1352e JG |
369 | /* In order to process neighbor discovery options, we need the |
370 | * entire packet. | |
371 | */ | |
372 | if (unlikely(icmp_len < sizeof(*nd))) | |
03f0d916 AZ |
373 | return 0; |
374 | ||
375 | if (unlikely(skb_linearize(skb))) | |
376 | return -ENOMEM; | |
ccb1352e JG |
377 | |
378 | nd = (struct nd_msg *)skb_transport_header(skb); | |
379 | key->ipv6.nd.target = nd->target; | |
ccb1352e JG |
380 | |
381 | icmp_len -= sizeof(*nd); | |
382 | offset = 0; | |
383 | while (icmp_len >= 8) { | |
384 | struct nd_opt_hdr *nd_opt = | |
385 | (struct nd_opt_hdr *)(nd->opt + offset); | |
386 | int opt_len = nd_opt->nd_opt_len * 8; | |
387 | ||
388 | if (unlikely(!opt_len || opt_len > icmp_len)) | |
03f0d916 | 389 | return 0; |
ccb1352e JG |
390 | |
391 | /* Store the link layer address if the appropriate | |
392 | * option is provided. It is considered an error if | |
393 | * the same link layer option is specified twice. | |
394 | */ | |
395 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR | |
396 | && opt_len == 8) { | |
397 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) | |
398 | goto invalid; | |
8c63ff09 JP |
399 | ether_addr_copy(key->ipv6.nd.sll, |
400 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
401 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
402 | && opt_len == 8) { | |
403 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) | |
404 | goto invalid; | |
8c63ff09 JP |
405 | ether_addr_copy(key->ipv6.nd.tll, |
406 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
407 | } |
408 | ||
409 | icmp_len -= opt_len; | |
410 | offset += opt_len; | |
411 | } | |
412 | } | |
413 | ||
03f0d916 | 414 | return 0; |
ccb1352e JG |
415 | |
416 | invalid: | |
417 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); | |
418 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
419 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
420 | ||
03f0d916 | 421 | return 0; |
ccb1352e JG |
422 | } |
423 | ||
424 | /** | |
83c8df26 | 425 | * key_extract - extracts a flow key from an Ethernet frame. |
ccb1352e JG |
426 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
427 | * Ethernet header | |
ccb1352e | 428 | * @key: output flow key |
ccb1352e JG |
429 | * |
430 | * The caller must ensure that skb->len >= ETH_HLEN. | |
431 | * | |
432 | * Returns 0 if successful, otherwise a negative errno value. | |
433 | * | |
434 | * Initializes @skb header pointers as follows: | |
435 | * | |
436 | * - skb->mac_header: the Ethernet header. | |
437 | * | |
438 | * - skb->network_header: just past the Ethernet header, or just past the | |
439 | * VLAN header, to the first byte of the Ethernet payload. | |
440 | * | |
34d94f21 | 441 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
ccb1352e JG |
442 | * on output, then just past the IP header, if one is present and |
443 | * of a correct length, otherwise the same as skb->network_header. | |
34d94f21 | 444 | * For other key->eth.type values it is left untouched. |
ccb1352e | 445 | */ |
83c8df26 | 446 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 447 | { |
03f0d916 | 448 | int error; |
ccb1352e JG |
449 | struct ethhdr *eth; |
450 | ||
f5796684 JG |
451 | /* Flags are always used as part of stats */ |
452 | key->tp.flags = 0; | |
453 | ||
ccb1352e JG |
454 | skb_reset_mac_header(skb); |
455 | ||
456 | /* Link layer. We are guaranteed to have at least the 14 byte Ethernet | |
457 | * header in the linear data area. | |
458 | */ | |
459 | eth = eth_hdr(skb); | |
8c63ff09 JP |
460 | ether_addr_copy(key->eth.src, eth->h_source); |
461 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
ccb1352e JG |
462 | |
463 | __skb_pull(skb, 2 * ETH_ALEN); | |
b34df5e8 PS |
464 | /* We are going to push all headers that we pull, so no need to |
465 | * update skb->csum here. | |
466 | */ | |
ccb1352e | 467 | |
07148121 | 468 | key->eth.tci = 0; |
ccb1352e JG |
469 | if (vlan_tx_tag_present(skb)) |
470 | key->eth.tci = htons(skb->vlan_tci); | |
471 | else if (eth->h_proto == htons(ETH_P_8021Q)) | |
472 | if (unlikely(parse_vlan(skb, key))) | |
473 | return -ENOMEM; | |
474 | ||
475 | key->eth.type = parse_ethertype(skb); | |
476 | if (unlikely(key->eth.type == htons(0))) | |
477 | return -ENOMEM; | |
478 | ||
479 | skb_reset_network_header(skb); | |
480 | __skb_push(skb, skb->data - skb_mac_header(skb)); | |
481 | ||
482 | /* Network layer. */ | |
483 | if (key->eth.type == htons(ETH_P_IP)) { | |
484 | struct iphdr *nh; | |
485 | __be16 offset; | |
486 | ||
ccb1352e JG |
487 | error = check_iphdr(skb); |
488 | if (unlikely(error)) { | |
07148121 JG |
489 | memset(&key->ip, 0, sizeof(key->ip)); |
490 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
491 | if (error == -EINVAL) { |
492 | skb->transport_header = skb->network_header; | |
493 | error = 0; | |
494 | } | |
03f0d916 | 495 | return error; |
ccb1352e JG |
496 | } |
497 | ||
498 | nh = ip_hdr(skb); | |
499 | key->ipv4.addr.src = nh->saddr; | |
500 | key->ipv4.addr.dst = nh->daddr; | |
501 | ||
502 | key->ip.proto = nh->protocol; | |
503 | key->ip.tos = nh->tos; | |
504 | key->ip.ttl = nh->ttl; | |
505 | ||
506 | offset = nh->frag_off & htons(IP_OFFSET); | |
507 | if (offset) { | |
508 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
03f0d916 | 509 | return 0; |
ccb1352e JG |
510 | } |
511 | if (nh->frag_off & htons(IP_MF) || | |
07148121 | 512 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
ccb1352e | 513 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
07148121 JG |
514 | else |
515 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
516 | |
517 | /* Transport layer. */ | |
518 | if (key->ip.proto == IPPROTO_TCP) { | |
ccb1352e JG |
519 | if (tcphdr_ok(skb)) { |
520 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
521 | key->tp.src = tcp->source; |
522 | key->tp.dst = tcp->dest; | |
523 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
524 | } else { |
525 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 526 | } |
07148121 | 527 | |
ccb1352e | 528 | } else if (key->ip.proto == IPPROTO_UDP) { |
ccb1352e JG |
529 | if (udphdr_ok(skb)) { |
530 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
531 | key->tp.src = udp->source; |
532 | key->tp.dst = udp->dest; | |
07148121 JG |
533 | } else { |
534 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 535 | } |
a175a723 JS |
536 | } else if (key->ip.proto == IPPROTO_SCTP) { |
537 | if (sctphdr_ok(skb)) { | |
538 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
539 | key->tp.src = sctp->source; |
540 | key->tp.dst = sctp->dest; | |
07148121 JG |
541 | } else { |
542 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 543 | } |
ccb1352e | 544 | } else if (key->ip.proto == IPPROTO_ICMP) { |
ccb1352e JG |
545 | if (icmphdr_ok(skb)) { |
546 | struct icmphdr *icmp = icmp_hdr(skb); | |
547 | /* The ICMP type and code fields use the 16-bit | |
548 | * transport port fields, so we need to store | |
549 | * them in 16-bit network byte order. */ | |
1139e241 JR |
550 | key->tp.src = htons(icmp->type); |
551 | key->tp.dst = htons(icmp->code); | |
07148121 JG |
552 | } else { |
553 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
554 | } |
555 | } | |
556 | ||
07148121 JG |
557 | } else if (key->eth.type == htons(ETH_P_ARP) || |
558 | key->eth.type == htons(ETH_P_RARP)) { | |
ccb1352e JG |
559 | struct arp_eth_header *arp; |
560 | ||
561 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
562 | ||
07148121 JG |
563 | if (arphdr_ok(skb) && |
564 | arp->ar_hrd == htons(ARPHRD_ETHER) && | |
565 | arp->ar_pro == htons(ETH_P_IP) && | |
566 | arp->ar_hln == ETH_ALEN && | |
567 | arp->ar_pln == 4) { | |
ccb1352e JG |
568 | |
569 | /* We only match on the lower 8 bits of the opcode. */ | |
570 | if (ntohs(arp->ar_op) <= 0xff) | |
571 | key->ip.proto = ntohs(arp->ar_op); | |
07148121 JG |
572 | else |
573 | key->ip.proto = 0; | |
574 | ||
d04d3829 MM |
575 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
576 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
8c63ff09 JP |
577 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
578 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
07148121 JG |
579 | } else { |
580 | memset(&key->ip, 0, sizeof(key->ip)); | |
581 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
582 | } |
583 | } else if (key->eth.type == htons(ETH_P_IPV6)) { | |
584 | int nh_len; /* IPv6 Header + Extensions */ | |
585 | ||
03f0d916 | 586 | nh_len = parse_ipv6hdr(skb, key); |
ccb1352e | 587 | if (unlikely(nh_len < 0)) { |
07148121 JG |
588 | memset(&key->ip, 0, sizeof(key->ip)); |
589 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
03f0d916 | 590 | if (nh_len == -EINVAL) { |
ccb1352e | 591 | skb->transport_header = skb->network_header; |
03f0d916 AZ |
592 | error = 0; |
593 | } else { | |
ccb1352e | 594 | error = nh_len; |
03f0d916 AZ |
595 | } |
596 | return error; | |
ccb1352e JG |
597 | } |
598 | ||
599 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) | |
03f0d916 | 600 | return 0; |
ccb1352e JG |
601 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
602 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
603 | ||
604 | /* Transport layer. */ | |
605 | if (key->ip.proto == NEXTHDR_TCP) { | |
ccb1352e JG |
606 | if (tcphdr_ok(skb)) { |
607 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
608 | key->tp.src = tcp->source; |
609 | key->tp.dst = tcp->dest; | |
610 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
611 | } else { |
612 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
613 | } |
614 | } else if (key->ip.proto == NEXTHDR_UDP) { | |
ccb1352e JG |
615 | if (udphdr_ok(skb)) { |
616 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
617 | key->tp.src = udp->source; |
618 | key->tp.dst = udp->dest; | |
07148121 JG |
619 | } else { |
620 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 621 | } |
a175a723 JS |
622 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
623 | if (sctphdr_ok(skb)) { | |
624 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
625 | key->tp.src = sctp->source; |
626 | key->tp.dst = sctp->dest; | |
07148121 JG |
627 | } else { |
628 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 629 | } |
ccb1352e | 630 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
ccb1352e | 631 | if (icmp6hdr_ok(skb)) { |
03f0d916 AZ |
632 | error = parse_icmpv6(skb, key, nh_len); |
633 | if (error) | |
634 | return error; | |
07148121 JG |
635 | } else { |
636 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
637 | } |
638 | } | |
639 | } | |
03f0d916 | 640 | return 0; |
ccb1352e | 641 | } |
83c8df26 | 642 | |
971427f3 AZ |
643 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
644 | { | |
645 | return key_extract(skb, key); | |
646 | } | |
647 | ||
f0b128c1 | 648 | int ovs_flow_key_extract(struct ovs_tunnel_info *tun_info, |
8c8b1b83 | 649 | struct sk_buff *skb, struct sw_flow_key *key) |
83c8df26 PS |
650 | { |
651 | /* Extract metadata from packet. */ | |
f5796684 | 652 | if (tun_info) { |
f0b128c1 | 653 | memcpy(&key->tun_key, &tun_info->tunnel, sizeof(key->tun_key)); |
f5796684 JG |
654 | |
655 | if (tun_info->options) { | |
656 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * | |
657 | 8)) - 1 | |
658 | > sizeof(key->tun_opts)); | |
659 | memcpy(GENEVE_OPTS(key, tun_info->options_len), | |
660 | tun_info->options, tun_info->options_len); | |
661 | key->tun_opts_len = tun_info->options_len; | |
662 | } else { | |
663 | key->tun_opts_len = 0; | |
664 | } | |
665 | } else { | |
666 | key->tun_opts_len = 0; | |
07148121 | 667 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
f5796684 | 668 | } |
83c8df26 PS |
669 | |
670 | key->phy.priority = skb->priority; | |
671 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
672 | key->phy.skb_mark = skb->mark; | |
07148121 JG |
673 | key->ovs_flow_hash = 0; |
674 | key->recirc_id = 0; | |
675 | ||
676 | /* Flags are always used as part of stats */ | |
677 | key->tp.flags = 0; | |
83c8df26 PS |
678 | |
679 | return key_extract(skb, key); | |
680 | } | |
681 | ||
682 | int ovs_flow_key_extract_userspace(const struct nlattr *attr, | |
683 | struct sk_buff *skb, | |
684 | struct sw_flow_key *key) | |
685 | { | |
686 | int err; | |
687 | ||
83c8df26 PS |
688 | /* Extract metadata from netlink attributes. */ |
689 | err = ovs_nla_get_flow_metadata(attr, key); | |
690 | if (err) | |
691 | return err; | |
692 | ||
693 | return key_extract(skb, key); | |
694 | } |