]>
Commit | Line | Data |
---|---|---|
064af421 | 1 | /* |
e23775f2 | 2 | * Copyright (c) 2007-2015 Nicira, Inc. |
a14bc59f | 3 | * |
a9a29d22 JG |
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 | |
064af421 BP |
17 | */ |
18 | ||
6455100f | 19 | #include <linux/uaccess.h> |
064af421 BP |
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> | |
8d5ebd83 | 26 | #include <linux/jhash.h> |
064af421 BP |
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> | |
a26ef517 | 32 | #include <linux/if_arp.h> |
064af421 | 33 | #include <linux/ip.h> |
d31f1109 | 34 | #include <linux/ipv6.h> |
2baf0e0c | 35 | #include <linux/mpls.h> |
10f72e3d | 36 | #include <linux/sctp.h> |
b0b906cc | 37 | #include <linux/smp.h> |
064af421 BP |
38 | #include <linux/tcp.h> |
39 | #include <linux/udp.h> | |
40 | #include <linux/icmp.h> | |
d31f1109 | 41 | #include <linux/icmpv6.h> |
3544358a | 42 | #include <linux/rculist.h> |
064af421 | 43 | #include <net/ip.h> |
d31f1109 | 44 | #include <net/ipv6.h> |
2baf0e0c | 45 | #include <net/mpls.h> |
685a51a5 | 46 | #include <net/ndisc.h> |
064af421 | 47 | |
c135bba1 | 48 | #include "datapath.h" |
a94ebc39 | 49 | #include "conntrack.h" |
c135bba1 PS |
50 | #include "flow.h" |
51 | #include "flow_netlink.h" | |
e23775f2 | 52 | #include "vport.h" |
6ce39213 JG |
53 | #include "vlan.h" |
54 | ||
a097c0b2 | 55 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
a1c564be | 56 | { |
a097c0b2 PS |
57 | struct timespec cur_ts; |
58 | u64 cur_ms, idle_ms; | |
a1c564be | 59 | |
a097c0b2 PS |
60 | ktime_get_ts(&cur_ts); |
61 | idle_ms = jiffies_to_msecs(jiffies - flow_jiffies); | |
62 | cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC + | |
63 | cur_ts.tv_nsec / NSEC_PER_MSEC; | |
a1c564be | 64 | |
a097c0b2 | 65 | return cur_ms - idle_ms; |
c2dd5e99 AZ |
66 | } |
67 | ||
a66733a8 | 68 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
a1c564be | 69 | |
a5b8d49b | 70 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
f1f60b85 | 71 | const struct sk_buff *skb) |
a1c564be | 72 | { |
b0f3a2fe | 73 | struct flow_stats *stats; |
9ac56358 | 74 | int node = numa_node_id(); |
efd8a18e | 75 | int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
a1c564be | 76 | |
9ac56358 | 77 | stats = rcu_dereference(flow->stats[node]); |
b0f3a2fe | 78 | |
9ac56358 JR |
79 | /* Check if already have node-specific stats. */ |
80 | if (likely(stats)) { | |
81 | spin_lock(&stats->lock); | |
82 | /* Mark if we write on the pre-allocated stats. */ | |
83 | if (node == 0 && unlikely(flow->stats_last_writer != node)) | |
84 | flow->stats_last_writer = node; | |
85 | } else { | |
86 | stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */ | |
87 | spin_lock(&stats->lock); | |
88 | ||
89 | /* If the current NUMA-node is the only writer on the | |
90 | * pre-allocated stats keep using them. | |
91 | */ | |
92 | if (unlikely(flow->stats_last_writer != node)) { | |
93 | /* A previous locker may have already allocated the | |
94 | * stats, so we need to check again. If node-specific | |
95 | * stats were already allocated, we update the pre- | |
96 | * allocated stats as we have already locked them. | |
97 | */ | |
98 | if (likely(flow->stats_last_writer != NUMA_NO_NODE) | |
bc36fcd3 | 99 | && likely(!rcu_access_pointer(flow->stats[node]))) { |
9ac56358 JR |
100 | /* Try to allocate node-specific stats. */ |
101 | struct flow_stats *new_stats; | |
102 | ||
103 | new_stats = | |
104 | kmem_cache_alloc_node(flow_stats_cache, | |
c0cddcec JS |
105 | GFP_NOWAIT | |
106 | __GFP_THISNODE | | |
107 | __GFP_NOWARN | | |
9ac56358 JR |
108 | __GFP_NOMEMALLOC, |
109 | node); | |
110 | if (likely(new_stats)) { | |
111 | new_stats->used = jiffies; | |
112 | new_stats->packet_count = 1; | |
9e917416 | 113 | new_stats->byte_count = len; |
9ac56358 JR |
114 | new_stats->tcp_flags = tcp_flags; |
115 | spin_lock_init(&new_stats->lock); | |
116 | ||
117 | rcu_assign_pointer(flow->stats[node], | |
118 | new_stats); | |
119 | goto unlock; | |
120 | } | |
121 | } | |
122 | flow->stats_last_writer = node; | |
123 | } | |
124 | } | |
125 | ||
b0b906cc PS |
126 | stats->used = jiffies; |
127 | stats->packet_count++; | |
9e917416 | 128 | stats->byte_count += len; |
b0b906cc | 129 | stats->tcp_flags |= tcp_flags; |
9ac56358 | 130 | unlock: |
b0b906cc PS |
131 | spin_unlock(&stats->lock); |
132 | } | |
133 | ||
4bb90bea JR |
134 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
135 | void ovs_flow_stats_get(const struct sw_flow *flow, | |
136 | struct ovs_flow_stats *ovs_stats, | |
b0f3a2fe PS |
137 | unsigned long *used, __be16 *tcp_flags) |
138 | { | |
9ac56358 | 139 | int node; |
b0b906cc | 140 | |
b0f3a2fe PS |
141 | *used = 0; |
142 | *tcp_flags = 0; | |
143 | memset(ovs_stats, 0, sizeof(*ovs_stats)); | |
b0b906cc | 144 | |
9ac56358 | 145 | for_each_node(node) { |
4bb90bea | 146 | struct flow_stats *stats = rcu_dereference_ovsl(flow->stats[node]); |
9ac56358 JR |
147 | |
148 | if (stats) { | |
149 | /* Local CPU may write on non-local stats, so we must | |
150 | * block bottom-halves here. | |
151 | */ | |
152 | spin_lock_bh(&stats->lock); | |
4029c043 | 153 | if (!*used || time_after(stats->used, *used)) |
9ac56358 JR |
154 | *used = stats->used; |
155 | *tcp_flags |= stats->tcp_flags; | |
156 | ovs_stats->n_packets += stats->packet_count; | |
157 | ovs_stats->n_bytes += stats->byte_count; | |
158 | spin_unlock_bh(&stats->lock); | |
159 | } | |
b0b906cc | 160 | } |
b0b906cc PS |
161 | } |
162 | ||
4bb90bea | 163 | /* Called with ovs_mutex. */ |
b0b906cc PS |
164 | void ovs_flow_stats_clear(struct sw_flow *flow) |
165 | { | |
9ac56358 JR |
166 | int node; |
167 | ||
168 | for_each_node(node) { | |
4bb90bea | 169 | struct flow_stats *stats = ovsl_dereference(flow->stats[node]); |
9ac56358 JR |
170 | |
171 | if (stats) { | |
172 | spin_lock_bh(&stats->lock); | |
173 | stats->used = 0; | |
174 | stats->packet_count = 0; | |
175 | stats->byte_count = 0; | |
176 | stats->tcp_flags = 0; | |
177 | spin_unlock_bh(&stats->lock); | |
178 | } | |
b0b906cc | 179 | } |
a1c564be AZ |
180 | } |
181 | ||
2db65bf7 JG |
182 | static int check_header(struct sk_buff *skb, int len) |
183 | { | |
184 | if (unlikely(skb->len < len)) | |
185 | return -EINVAL; | |
186 | if (unlikely(!pskb_may_pull(skb, len))) | |
187 | return -ENOMEM; | |
188 | return 0; | |
189 | } | |
190 | ||
6455100f | 191 | static bool arphdr_ok(struct sk_buff *skb) |
a26ef517 | 192 | { |
2db65bf7 JG |
193 | return pskb_may_pull(skb, skb_network_offset(skb) + |
194 | sizeof(struct arp_eth_header)); | |
a26ef517 JP |
195 | } |
196 | ||
6455100f | 197 | static int check_iphdr(struct sk_buff *skb) |
064af421 | 198 | { |
4c1ad233 BP |
199 | unsigned int nh_ofs = skb_network_offset(skb); |
200 | unsigned int ip_len; | |
2db65bf7 | 201 | int err; |
4c1ad233 | 202 | |
2db65bf7 JG |
203 | err = check_header(skb, nh_ofs + sizeof(struct iphdr)); |
204 | if (unlikely(err)) | |
205 | return err; | |
4c1ad233 BP |
206 | |
207 | ip_len = ip_hdrlen(skb); | |
2db65bf7 JG |
208 | if (unlikely(ip_len < sizeof(struct iphdr) || |
209 | skb->len < nh_ofs + ip_len)) | |
4c1ad233 BP |
210 | return -EINVAL; |
211 | ||
4c1ad233 BP |
212 | skb_set_transport_header(skb, nh_ofs + ip_len); |
213 | return 0; | |
064af421 BP |
214 | } |
215 | ||
6455100f | 216 | static bool tcphdr_ok(struct sk_buff *skb) |
064af421 BP |
217 | { |
218 | int th_ofs = skb_transport_offset(skb); | |
2db65bf7 JG |
219 | int tcp_len; |
220 | ||
221 | if (unlikely(!pskb_may_pull(skb, th_ofs + sizeof(struct tcphdr)))) | |
222 | return false; | |
223 | ||
224 | tcp_len = tcp_hdrlen(skb); | |
225 | if (unlikely(tcp_len < sizeof(struct tcphdr) || | |
226 | skb->len < th_ofs + tcp_len)) | |
227 | return false; | |
228 | ||
229 | return true; | |
064af421 BP |
230 | } |
231 | ||
6455100f | 232 | static bool udphdr_ok(struct sk_buff *skb) |
064af421 | 233 | { |
2db65bf7 JG |
234 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
235 | sizeof(struct udphdr)); | |
064af421 BP |
236 | } |
237 | ||
10f72e3d JS |
238 | static bool sctphdr_ok(struct sk_buff *skb) |
239 | { | |
240 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
241 | sizeof(struct sctphdr)); | |
242 | } | |
243 | ||
6455100f | 244 | static bool icmphdr_ok(struct sk_buff *skb) |
064af421 | 245 | { |
2db65bf7 JG |
246 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
247 | sizeof(struct icmphdr)); | |
064af421 BP |
248 | } |
249 | ||
a1c564be | 250 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
d31f1109 JP |
251 | { |
252 | unsigned int nh_ofs = skb_network_offset(skb); | |
253 | unsigned int nh_len; | |
254 | int payload_ofs; | |
d31f1109 JP |
255 | struct ipv6hdr *nh; |
256 | uint8_t nexthdr; | |
0fd0d083 | 257 | __be16 frag_off; |
2db65bf7 | 258 | int err; |
d31f1109 | 259 | |
2db65bf7 JG |
260 | err = check_header(skb, nh_ofs + sizeof(*nh)); |
261 | if (unlikely(err)) | |
262 | return err; | |
d31f1109 JP |
263 | |
264 | nh = ipv6_hdr(skb); | |
265 | nexthdr = nh->nexthdr; | |
266 | payload_ofs = (u8 *)(nh + 1) - skb->data; | |
d31f1109 | 267 | |
28bad473 | 268 | key->ip.proto = NEXTHDR_NONE; |
530180fd | 269 | key->ip.tos = ipv6_get_dsfield(nh); |
a61680c6 | 270 | key->ip.ttl = nh->hop_limit; |
fa8223b7 | 271 | key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); |
858d1026 JG |
272 | key->ipv6.addr.src = nh->saddr; |
273 | key->ipv6.addr.dst = nh->daddr; | |
d31f1109 | 274 | |
0fd0d083 | 275 | payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); |
bfef4717 | 276 | if (unlikely(payload_ofs < 0)) |
d31f1109 | 277 | return -EINVAL; |
bfef4717 | 278 | |
0fd0d083 JG |
279 | if (frag_off) { |
280 | if (frag_off & htons(~0x7)) | |
281 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
282 | else | |
283 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
9cef26ac JG |
284 | } else { |
285 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
0fd0d083 JG |
286 | } |
287 | ||
d31f1109 | 288 | nh_len = payload_ofs - nh_ofs; |
d31f1109 | 289 | skb_set_transport_header(skb, nh_ofs + nh_len); |
28bad473 | 290 | key->ip.proto = nexthdr; |
d31f1109 JP |
291 | return nh_len; |
292 | } | |
293 | ||
294 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
295 | { | |
2db65bf7 JG |
296 | return pskb_may_pull(skb, skb_transport_offset(skb) + |
297 | sizeof(struct icmp6hdr)); | |
d31f1109 | 298 | } |
ec58547a | 299 | |
2db65bf7 | 300 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) |
064af421 | 301 | { |
50f06e16 BP |
302 | struct qtag_prefix { |
303 | __be16 eth_type; /* ETH_P_8021Q */ | |
304 | __be16 tci; | |
305 | }; | |
306 | struct qtag_prefix *qp; | |
307 | ||
8ddc056d BP |
308 | if (unlikely(skb->len < sizeof(struct qtag_prefix) + sizeof(__be16))) |
309 | return 0; | |
310 | ||
2db65bf7 JG |
311 | if (unlikely(!pskb_may_pull(skb, sizeof(struct qtag_prefix) + |
312 | sizeof(__be16)))) | |
313 | return -ENOMEM; | |
50f06e16 BP |
314 | |
315 | qp = (struct qtag_prefix *) skb->data; | |
76abe283 | 316 | key->eth.tci = qp->tci | htons(VLAN_TAG_PRESENT); |
50f06e16 | 317 | __skb_pull(skb, sizeof(struct qtag_prefix)); |
2db65bf7 JG |
318 | |
319 | return 0; | |
50f06e16 BP |
320 | } |
321 | ||
322 | static __be16 parse_ethertype(struct sk_buff *skb) | |
064af421 | 323 | { |
50f06e16 BP |
324 | struct llc_snap_hdr { |
325 | u8 dsap; /* Always 0xAA */ | |
326 | u8 ssap; /* Always 0xAA */ | |
327 | u8 ctrl; | |
328 | u8 oui[3]; | |
8dda8c9b | 329 | __be16 ethertype; |
50f06e16 BP |
330 | }; |
331 | struct llc_snap_hdr *llc; | |
332 | __be16 proto; | |
333 | ||
334 | proto = *(__be16 *) skb->data; | |
335 | __skb_pull(skb, sizeof(__be16)); | |
336 | ||
935fc582 | 337 | if (eth_proto_is_802_3(proto)) |
50f06e16 BP |
338 | return proto; |
339 | ||
2db65bf7 | 340 | if (skb->len < sizeof(struct llc_snap_hdr)) |
36956a7d | 341 | return htons(ETH_P_802_2); |
50f06e16 | 342 | |
2db65bf7 JG |
343 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) |
344 | return htons(0); | |
345 | ||
50f06e16 BP |
346 | llc = (struct llc_snap_hdr *) skb->data; |
347 | if (llc->dsap != LLC_SAP_SNAP || | |
348 | llc->ssap != LLC_SAP_SNAP || | |
349 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
36956a7d | 350 | return htons(ETH_P_802_2); |
50f06e16 BP |
351 | |
352 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
9e69bc5f | 353 | |
935fc582 | 354 | if (eth_proto_is_802_3(llc->ethertype)) |
9e69bc5f RL |
355 | return llc->ethertype; |
356 | ||
357 | return htons(ETH_P_802_2); | |
064af421 BP |
358 | } |
359 | ||
685a51a5 | 360 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, |
a1c564be | 361 | int nh_len) |
685a51a5 | 362 | { |
685a51a5 JP |
363 | struct icmp6hdr *icmp = icmp6_hdr(skb); |
364 | ||
365 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
bfef4717 JG |
366 | * fields, so we need to store them in 16-bit network byte order. |
367 | */ | |
708fb4c5 JR |
368 | key->tp.src = htons(icmp->icmp6_type); |
369 | key->tp.dst = htons(icmp->icmp6_code); | |
9cef26ac | 370 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
685a51a5 | 371 | |
bfef4717 JG |
372 | if (icmp->icmp6_code == 0 && |
373 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
374 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
e977ba19 | 375 | int icmp_len = skb->len - skb_transport_offset(skb); |
685a51a5 JP |
376 | struct nd_msg *nd; |
377 | int offset; | |
378 | ||
379 | /* In order to process neighbor discovery options, we need the | |
bfef4717 JG |
380 | * entire packet. |
381 | */ | |
382 | if (unlikely(icmp_len < sizeof(*nd))) | |
a1c564be AZ |
383 | return 0; |
384 | ||
385 | if (unlikely(skb_linearize(skb))) | |
386 | return -ENOMEM; | |
685a51a5 JP |
387 | |
388 | nd = (struct nd_msg *)skb_transport_header(skb); | |
858d1026 | 389 | key->ipv6.nd.target = nd->target; |
685a51a5 JP |
390 | |
391 | icmp_len -= sizeof(*nd); | |
392 | offset = 0; | |
393 | while (icmp_len >= 8) { | |
6455100f PS |
394 | struct nd_opt_hdr *nd_opt = |
395 | (struct nd_opt_hdr *)(nd->opt + offset); | |
685a51a5 JP |
396 | int opt_len = nd_opt->nd_opt_len * 8; |
397 | ||
bfef4717 | 398 | if (unlikely(!opt_len || opt_len > icmp_len)) |
a1c564be | 399 | return 0; |
685a51a5 | 400 | |
bfef4717 JG |
401 | /* Store the link layer address if the appropriate |
402 | * option is provided. It is considered an error if | |
403 | * the same link layer option is specified twice. | |
404 | */ | |
685a51a5 | 405 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR |
bfef4717 | 406 | && opt_len == 8) { |
76abe283 | 407 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) |
685a51a5 | 408 | goto invalid; |
982a47ec | 409 | ether_addr_copy(key->ipv6.nd.sll, |
7d16c847 | 410 | &nd->opt[offset+sizeof(*nd_opt)]); |
685a51a5 | 411 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
bfef4717 | 412 | && opt_len == 8) { |
76abe283 | 413 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) |
685a51a5 | 414 | goto invalid; |
982a47ec | 415 | ether_addr_copy(key->ipv6.nd.tll, |
7d16c847 | 416 | &nd->opt[offset+sizeof(*nd_opt)]); |
685a51a5 JP |
417 | } |
418 | ||
419 | icmp_len -= opt_len; | |
420 | offset += opt_len; | |
421 | } | |
422 | } | |
423 | ||
a1c564be | 424 | return 0; |
685a51a5 JP |
425 | |
426 | invalid: | |
76abe283 AE |
427 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); |
428 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
429 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
685a51a5 | 430 | |
a1c564be | 431 | return 0; |
685a51a5 JP |
432 | } |
433 | ||
a31e0e31 | 434 | /** |
c135bba1 | 435 | * key_extract - extracts a flow key from an Ethernet frame. |
a31e0e31 BP |
436 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
437 | * Ethernet header | |
a31e0e31 BP |
438 | * @key: output flow key |
439 | * | |
440 | * The caller must ensure that skb->len >= ETH_HLEN. | |
441 | * | |
4c1ad233 BP |
442 | * Returns 0 if successful, otherwise a negative errno value. |
443 | * | |
59a18f80 BP |
444 | * Initializes @skb header pointers as follows: |
445 | * | |
446 | * - skb->mac_header: the Ethernet header. | |
447 | * | |
448 | * - skb->network_header: just past the Ethernet header, or just past the | |
449 | * VLAN header, to the first byte of the Ethernet payload. | |
450 | * | |
ae9020cf | 451 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
d31f1109 JP |
452 | * on output, then just past the IP header, if one is present and |
453 | * of a correct length, otherwise the same as skb->network_header. | |
ae9020cf | 454 | * For other key->eth.type values it is left untouched. |
a31e0e31 | 455 | */ |
c135bba1 | 456 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
064af421 | 457 | { |
a1c564be | 458 | int error; |
064af421 | 459 | struct ethhdr *eth; |
064af421 | 460 | |
7d16c847 | 461 | /* Flags are always used as part of stats */ |
9cef26ac | 462 | key->tp.flags = 0; |
064af421 | 463 | |
064af421 | 464 | skb_reset_mac_header(skb); |
064af421 | 465 | |
2db65bf7 JG |
466 | /* Link layer. We are guaranteed to have at least the 14 byte Ethernet |
467 | * header in the linear data area. | |
468 | */ | |
50f06e16 | 469 | eth = eth_hdr(skb); |
982a47ec JP |
470 | ether_addr_copy(key->eth.src, eth->h_source); |
471 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
76abe283 | 472 | |
50f06e16 | 473 | __skb_pull(skb, 2 * ETH_ALEN); |
3cfede14 | 474 | /* We are going to push all headers that we pull, so no need to |
af465b67 PS |
475 | * update skb->csum here. |
476 | */ | |
6ce39213 | 477 | |
9cef26ac | 478 | key->eth.tci = 0; |
efd8a18e | 479 | if (skb_vlan_tag_present(skb)) |
76abe283 | 480 | key->eth.tci = htons(vlan_get_tci(skb)); |
6ce39213 | 481 | else if (eth->h_proto == htons(ETH_P_8021Q)) |
2db65bf7 JG |
482 | if (unlikely(parse_vlan(skb, key))) |
483 | return -ENOMEM; | |
6ce39213 | 484 | |
76abe283 AE |
485 | key->eth.type = parse_ethertype(skb); |
486 | if (unlikely(key->eth.type == htons(0))) | |
2db65bf7 JG |
487 | return -ENOMEM; |
488 | ||
50f06e16 | 489 | skb_reset_network_header(skb); |
ccf43786 | 490 | skb_reset_mac_len(skb); |
2db65bf7 | 491 | __skb_push(skb, skb->data - skb_mac_header(skb)); |
064af421 BP |
492 | |
493 | /* Network layer. */ | |
76abe283 | 494 | if (key->eth.type == htons(ETH_P_IP)) { |
4c1ad233 | 495 | struct iphdr *nh; |
7257b535 | 496 | __be16 offset; |
76abe283 | 497 | |
4c1ad233 BP |
498 | error = check_iphdr(skb); |
499 | if (unlikely(error)) { | |
9cef26ac JG |
500 | memset(&key->ip, 0, sizeof(key->ip)); |
501 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
4c1ad233 BP |
502 | if (error == -EINVAL) { |
503 | skb->transport_header = skb->network_header; | |
76abe283 | 504 | error = 0; |
4c1ad233 | 505 | } |
a1c564be | 506 | return error; |
4c1ad233 BP |
507 | } |
508 | ||
509 | nh = ip_hdr(skb); | |
76abe283 AE |
510 | key->ipv4.addr.src = nh->saddr; |
511 | key->ipv4.addr.dst = nh->daddr; | |
7257b535 | 512 | |
28bad473 | 513 | key->ip.proto = nh->protocol; |
530180fd | 514 | key->ip.tos = nh->tos; |
a61680c6 | 515 | key->ip.ttl = nh->ttl; |
064af421 | 516 | |
7257b535 BP |
517 | offset = nh->frag_off & htons(IP_OFFSET); |
518 | if (offset) { | |
9e44d715 | 519 | key->ip.frag = OVS_FRAG_TYPE_LATER; |
a1c564be | 520 | return 0; |
7257b535 BP |
521 | } |
522 | if (nh->frag_off & htons(IP_MF) || | |
7d16c847 | 523 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
9e44d715 | 524 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
9cef26ac JG |
525 | else |
526 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
b7a31ec1 | 527 | |
7257b535 | 528 | /* Transport layer. */ |
28bad473 | 529 | if (key->ip.proto == IPPROTO_TCP) { |
7257b535 | 530 | if (tcphdr_ok(skb)) { |
5f4d087c | 531 | struct tcphdr *tcp = tcp_hdr(skb); |
708fb4c5 JR |
532 | key->tp.src = tcp->source; |
533 | key->tp.dst = tcp->dest; | |
534 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
9cef26ac JG |
535 | } else { |
536 | memset(&key->tp, 0, sizeof(key->tp)); | |
5f4d087c | 537 | } |
7d16c847 | 538 | |
28bad473 | 539 | } else if (key->ip.proto == IPPROTO_UDP) { |
7257b535 | 540 | if (udphdr_ok(skb)) { |
5f4d087c | 541 | struct udphdr *udp = udp_hdr(skb); |
708fb4c5 JR |
542 | key->tp.src = udp->source; |
543 | key->tp.dst = udp->dest; | |
9cef26ac JG |
544 | } else { |
545 | memset(&key->tp, 0, sizeof(key->tp)); | |
5f4d087c | 546 | } |
10f72e3d JS |
547 | } else if (key->ip.proto == IPPROTO_SCTP) { |
548 | if (sctphdr_ok(skb)) { | |
549 | struct sctphdr *sctp = sctp_hdr(skb); | |
708fb4c5 JR |
550 | key->tp.src = sctp->source; |
551 | key->tp.dst = sctp->dest; | |
9cef26ac JG |
552 | } else { |
553 | memset(&key->tp, 0, sizeof(key->tp)); | |
10f72e3d | 554 | } |
28bad473 | 555 | } else if (key->ip.proto == IPPROTO_ICMP) { |
7257b535 | 556 | if (icmphdr_ok(skb)) { |
5f4d087c JG |
557 | struct icmphdr *icmp = icmp_hdr(skb); |
558 | /* The ICMP type and code fields use the 16-bit | |
6455100f | 559 | * transport port fields, so we need to store |
af465b67 PS |
560 | * them in 16-bit network byte order. |
561 | */ | |
708fb4c5 JR |
562 | key->tp.src = htons(icmp->type); |
563 | key->tp.dst = htons(icmp->code); | |
9cef26ac JG |
564 | } else { |
565 | memset(&key->tp, 0, sizeof(key->tp)); | |
5f4d087c JG |
566 | } |
567 | } | |
568 | ||
9cef26ac JG |
569 | } else if (key->eth.type == htons(ETH_P_ARP) || |
570 | key->eth.type == htons(ETH_P_RARP)) { | |
a26ef517 | 571 | struct arp_eth_header *arp; |
10173cea | 572 | bool arp_available = arphdr_ok(skb); |
a26ef517 JP |
573 | |
574 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
575 | ||
10173cea LR |
576 | if (arp_available && |
577 | arp->ar_hrd == htons(ARPHRD_ETHER) && | |
578 | arp->ar_pro == htons(ETH_P_IP) && | |
579 | arp->ar_hln == ETH_ALEN && | |
580 | arp->ar_pln == 4) { | |
de3f65ea JP |
581 | |
582 | /* We only match on the lower 8 bits of the opcode. */ | |
b7a31ec1 | 583 | if (ntohs(arp->ar_op) <= 0xff) |
28bad473 | 584 | key->ip.proto = ntohs(arp->ar_op); |
9cef26ac JG |
585 | else |
586 | key->ip.proto = 0; | |
587 | ||
a3d3ad0c MM |
588 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
589 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
982a47ec JP |
590 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
591 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
9cef26ac JG |
592 | } else { |
593 | memset(&key->ip, 0, sizeof(key->ip)); | |
594 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
de3f65ea | 595 | } |
ccf43786 SH |
596 | } else if (eth_p_mpls(key->eth.type)) { |
597 | size_t stack_len = MPLS_HLEN; | |
598 | ||
599 | /* In the presence of an MPLS label stack the end of the L2 | |
600 | * header and the beginning of the L3 header differ. | |
601 | * | |
602 | * Advance network_header to the beginning of the L3 | |
603 | * header. mac_len corresponds to the end of the L2 header. | |
604 | */ | |
605 | while (1) { | |
606 | __be32 lse; | |
607 | ||
608 | error = check_header(skb, skb->mac_len + stack_len); | |
609 | if (unlikely(error)) | |
610 | return 0; | |
611 | ||
612 | memcpy(&lse, skb_network_header(skb), MPLS_HLEN); | |
613 | ||
614 | if (stack_len == MPLS_HLEN) | |
615 | memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN); | |
616 | ||
617 | skb_set_network_header(skb, skb->mac_len + stack_len); | |
2baf0e0c | 618 | if (lse & htonl(MPLS_LS_S_MASK)) |
ccf43786 SH |
619 | break; |
620 | ||
621 | stack_len += MPLS_HLEN; | |
622 | } | |
76abe283 | 623 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
d31f1109 JP |
624 | int nh_len; /* IPv6 Header + Extensions */ |
625 | ||
a1c564be | 626 | nh_len = parse_ipv6hdr(skb, key); |
d31f1109 | 627 | if (unlikely(nh_len < 0)) { |
9cef26ac JG |
628 | memset(&key->ip, 0, sizeof(key->ip)); |
629 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
a1c564be | 630 | if (nh_len == -EINVAL) { |
d31f1109 | 631 | skb->transport_header = skb->network_header; |
a1c564be AZ |
632 | error = 0; |
633 | } else { | |
76abe283 | 634 | error = nh_len; |
a1c564be AZ |
635 | } |
636 | return error; | |
d31f1109 JP |
637 | } |
638 | ||
9e44d715 | 639 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) |
a1c564be | 640 | return 0; |
7257b535 | 641 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
9e44d715 | 642 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
7257b535 | 643 | |
d31f1109 | 644 | /* Transport layer. */ |
28bad473 | 645 | if (key->ip.proto == NEXTHDR_TCP) { |
d31f1109 JP |
646 | if (tcphdr_ok(skb)) { |
647 | struct tcphdr *tcp = tcp_hdr(skb); | |
708fb4c5 JR |
648 | key->tp.src = tcp->source; |
649 | key->tp.dst = tcp->dest; | |
650 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
9cef26ac JG |
651 | } else { |
652 | memset(&key->tp, 0, sizeof(key->tp)); | |
d31f1109 | 653 | } |
28bad473 | 654 | } else if (key->ip.proto == NEXTHDR_UDP) { |
d31f1109 JP |
655 | if (udphdr_ok(skb)) { |
656 | struct udphdr *udp = udp_hdr(skb); | |
708fb4c5 JR |
657 | key->tp.src = udp->source; |
658 | key->tp.dst = udp->dest; | |
9cef26ac JG |
659 | } else { |
660 | memset(&key->tp, 0, sizeof(key->tp)); | |
d31f1109 | 661 | } |
10f72e3d JS |
662 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
663 | if (sctphdr_ok(skb)) { | |
664 | struct sctphdr *sctp = sctp_hdr(skb); | |
708fb4c5 JR |
665 | key->tp.src = sctp->source; |
666 | key->tp.dst = sctp->dest; | |
9cef26ac JG |
667 | } else { |
668 | memset(&key->tp, 0, sizeof(key->tp)); | |
10f72e3d | 669 | } |
28bad473 | 670 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
d31f1109 | 671 | if (icmp6hdr_ok(skb)) { |
a1c564be AZ |
672 | error = parse_icmpv6(skb, key, nh_len); |
673 | if (error) | |
674 | return error; | |
9cef26ac JG |
675 | } else { |
676 | memset(&key->tp, 0, sizeof(key->tp)); | |
d31f1109 JP |
677 | } |
678 | } | |
064af421 | 679 | } |
a1c564be | 680 | return 0; |
064af421 | 681 | } |
c135bba1 | 682 | |
e0a42ae2 AZ |
683 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
684 | { | |
685 | return key_extract(skb, key); | |
686 | } | |
687 | ||
e23775f2 | 688 | int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info, |
7d16c847 | 689 | struct sk_buff *skb, struct sw_flow_key *key) |
c135bba1 PS |
690 | { |
691 | /* Extract metadata from packet. */ | |
fb66fbd1 | 692 | if (tun_info) { |
e23775f2 PS |
693 | if (ip_tunnel_info_af(tun_info) != AF_INET) |
694 | return -EINVAL; | |
c135bba1 | 695 | |
e23775f2 | 696 | memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key)); |
c135bba1 PS |
697 | BUILD_BUG_ON(((1 << (sizeof(tun_info->options_len) * 8)) - 1) > |
698 | sizeof(key->tun_opts)); | |
699 | ||
e23775f2 PS |
700 | if (tun_info->options_len) { |
701 | ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len), | |
702 | tun_info); | |
c135bba1 PS |
703 | key->tun_opts_len = tun_info->options_len; |
704 | } else { | |
705 | key->tun_opts_len = 0; | |
706 | } | |
707 | } else { | |
708 | key->tun_opts_len = 0; | |
709 | memset(&key->tun_key, 0, sizeof(key->tun_key)); | |
710 | } | |
711 | ||
712 | key->phy.priority = skb->priority; | |
713 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
714 | key->phy.skb_mark = skb->mark; | |
a94ebc39 | 715 | ovs_ct_fill_key(skb, key); |
c135bba1 PS |
716 | key->ovs_flow_hash = 0; |
717 | key->recirc_id = 0; | |
718 | ||
719 | return key_extract(skb, key); | |
720 | } | |
721 | ||
038e34ab | 722 | int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr, |
c135bba1 | 723 | struct sk_buff *skb, |
9233cef7 | 724 | struct sw_flow_key *key, bool log) |
c135bba1 PS |
725 | { |
726 | int err; | |
727 | ||
728 | /* Extract metadata from netlink attributes. */ | |
038e34ab | 729 | err = ovs_nla_get_flow_metadata(net, attr, key, log); |
c135bba1 PS |
730 | if (err) |
731 | return err; | |
732 | ||
733 | return key_extract(skb, key); | |
734 | } |