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> | |
db74a333 | 32 | #include <linux/cpumask.h> |
ccb1352e | 33 | #include <linux/if_arp.h> |
ccb1352e JG |
34 | #include <linux/ip.h> |
35 | #include <linux/ipv6.h> | |
25cd9ba0 | 36 | #include <linux/mpls.h> |
a175a723 | 37 | #include <linux/sctp.h> |
e298e505 | 38 | #include <linux/smp.h> |
ccb1352e JG |
39 | #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> | |
7d5437c7 | 45 | #include <net/ip_tunnels.h> |
ccb1352e | 46 | #include <net/ipv6.h> |
25cd9ba0 | 47 | #include <net/mpls.h> |
ccb1352e JG |
48 | #include <net/ndisc.h> |
49 | ||
a581b96d | 50 | #include "conntrack.h" |
83c8df26 PS |
51 | #include "datapath.h" |
52 | #include "flow.h" | |
53 | #include "flow_netlink.h" | |
a581b96d | 54 | #include "vport.h" |
83c8df26 | 55 | |
e6445719 | 56 | u64 ovs_flow_used_time(unsigned long flow_jiffies) |
03f0d916 | 57 | { |
e6445719 PS |
58 | struct timespec cur_ts; |
59 | u64 cur_ms, idle_ms; | |
03f0d916 | 60 | |
e6445719 PS |
61 | 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; | |
03f0d916 | 65 | |
e6445719 | 66 | return cur_ms - idle_ms; |
5828cd9a AZ |
67 | } |
68 | ||
df23e9f6 | 69 | #define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF)) |
03f0d916 | 70 | |
ad552007 | 71 | void ovs_flow_stats_update(struct sw_flow *flow, __be16 tcp_flags, |
12eb18f7 | 72 | const struct sk_buff *skb) |
03f0d916 | 73 | { |
e298e505 | 74 | struct flow_stats *stats; |
63e7959c | 75 | int node = numa_node_id(); |
db74a333 | 76 | int cpu = smp_processor_id(); |
df8a39de | 77 | int len = skb->len + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
03f0d916 | 78 | |
db74a333 | 79 | stats = rcu_dereference(flow->stats[cpu]); |
e298e505 | 80 | |
db74a333 | 81 | /* Check if already have CPU-specific stats. */ |
63e7959c JR |
82 | if (likely(stats)) { |
83 | spin_lock(&stats->lock); | |
84 | /* Mark if we write on the pre-allocated stats. */ | |
db74a333 TLSC |
85 | if (cpu == 0 && unlikely(flow->stats_last_writer != cpu)) |
86 | flow->stats_last_writer = cpu; | |
63e7959c JR |
87 | } else { |
88 | stats = rcu_dereference(flow->stats[0]); /* Pre-allocated. */ | |
89 | spin_lock(&stats->lock); | |
90 | ||
db74a333 | 91 | /* If the current CPU is the only writer on the |
63e7959c JR |
92 | * pre-allocated stats keep using them. |
93 | */ | |
db74a333 | 94 | if (unlikely(flow->stats_last_writer != cpu)) { |
63e7959c | 95 | /* A previous locker may have already allocated the |
db74a333 | 96 | * stats, so we need to check again. If CPU-specific |
63e7959c JR |
97 | * stats were already allocated, we update the pre- |
98 | * allocated stats as we have already locked them. | |
99 | */ | |
db74a333 TLSC |
100 | if (likely(flow->stats_last_writer != -1) && |
101 | likely(!rcu_access_pointer(flow->stats[cpu]))) { | |
102 | /* Try to allocate CPU-specific stats. */ | |
63e7959c JR |
103 | struct flow_stats *new_stats; |
104 | ||
105 | new_stats = | |
106 | kmem_cache_alloc_node(flow_stats_cache, | |
4167e9b2 DR |
107 | GFP_NOWAIT | |
108 | __GFP_THISNODE | | |
109 | __GFP_NOWARN | | |
63e7959c JR |
110 | __GFP_NOMEMALLOC, |
111 | node); | |
112 | if (likely(new_stats)) { | |
113 | new_stats->used = jiffies; | |
114 | new_stats->packet_count = 1; | |
24cc59d1 | 115 | new_stats->byte_count = len; |
63e7959c JR |
116 | new_stats->tcp_flags = tcp_flags; |
117 | spin_lock_init(&new_stats->lock); | |
118 | ||
db74a333 | 119 | rcu_assign_pointer(flow->stats[cpu], |
63e7959c JR |
120 | new_stats); |
121 | goto unlock; | |
122 | } | |
123 | } | |
db74a333 | 124 | flow->stats_last_writer = cpu; |
63e7959c JR |
125 | } |
126 | } | |
127 | ||
e298e505 PS |
128 | stats->used = jiffies; |
129 | stats->packet_count++; | |
24cc59d1 | 130 | stats->byte_count += len; |
e298e505 | 131 | stats->tcp_flags |= tcp_flags; |
63e7959c | 132 | unlock: |
e298e505 PS |
133 | spin_unlock(&stats->lock); |
134 | } | |
135 | ||
86ec8dba JR |
136 | /* Must be called with rcu_read_lock or ovs_mutex. */ |
137 | void ovs_flow_stats_get(const struct sw_flow *flow, | |
138 | struct ovs_flow_stats *ovs_stats, | |
e298e505 PS |
139 | unsigned long *used, __be16 *tcp_flags) |
140 | { | |
db74a333 | 141 | int cpu; |
e298e505 PS |
142 | |
143 | *used = 0; | |
144 | *tcp_flags = 0; | |
145 | memset(ovs_stats, 0, sizeof(*ovs_stats)); | |
146 | ||
db74a333 TLSC |
147 | /* 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]); | |
63e7959c JR |
150 | |
151 | 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 | } | |
e298e505 | 163 | } |
e298e505 PS |
164 | } |
165 | ||
86ec8dba | 166 | /* Called with ovs_mutex. */ |
e298e505 PS |
167 | void ovs_flow_stats_clear(struct sw_flow *flow) |
168 | { | |
db74a333 | 169 | int cpu; |
63e7959c | 170 | |
db74a333 TLSC |
171 | /* 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]); | |
63e7959c JR |
174 | |
175 | 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 | } | |
03f0d916 AZ |
184 | } |
185 | ||
ccb1352e JG |
186 | static 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 | ||
195 | static 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 | ||
201 | static 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 | ||
220 | static 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 | ||
236 | static bool udphdr_ok(struct sk_buff *skb) | |
237 | { | |
238 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
239 | sizeof(struct udphdr)); | |
240 | } | |
241 | ||
a175a723 JS |
242 | static bool sctphdr_ok(struct sk_buff *skb) |
243 | { | |
244 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
245 | sizeof(struct sctphdr)); | |
246 | } | |
247 | ||
ccb1352e JG |
248 | static bool icmphdr_ok(struct sk_buff *skb) |
249 | { | |
250 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
251 | sizeof(struct icmphdr)); | |
252 | } | |
253 | ||
03f0d916 | 254 | static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e JG |
255 | { |
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 | ||
ccb1352e JG |
264 | 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); | |
ccb1352e JG |
280 | |
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; | |
25ef1328 PS |
286 | } else { |
287 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
288 | } |
289 | ||
c30da497 SH |
290 | /* 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 | ||
ccb1352e JG |
297 | 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 | ||
303 | static bool icmp6hdr_ok(struct sk_buff *skb) | |
304 | { | |
305 | return pskb_may_pull(skb, skb_transport_offset(skb) + | |
306 | sizeof(struct icmp6hdr)); | |
307 | } | |
308 | ||
018c1dda EG |
309 | /** |
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 | */ | |
315 | static int parse_vlan_tag(struct sk_buff *skb, struct vlan_head *key_vh) | |
ccb1352e | 316 | { |
018c1dda | 317 | struct vlan_head *vh = (struct vlan_head *)skb->data; |
ccb1352e | 318 | |
018c1dda | 319 | if (likely(!eth_type_vlan(vh->tpid))) |
ccb1352e JG |
320 | return 0; |
321 | ||
018c1dda EG |
322 | 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)))) | |
ccb1352e JG |
327 | return -ENOMEM; |
328 | ||
018c1dda EG |
329 | 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 | ||
5108bbad | 337 | static void clear_vlan(struct sw_flow_key *key) |
018c1dda | 338 | { |
018c1dda EG |
339 | key->eth.vlan.tci = 0; |
340 | key->eth.vlan.tpid = 0; | |
341 | key->eth.cvlan.tci = 0; | |
342 | key->eth.cvlan.tpid = 0; | |
5108bbad JB |
343 | } |
344 | ||
345 | static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) | |
346 | { | |
347 | int res; | |
018c1dda | 348 | |
20ecf1e4 | 349 | if (skb_vlan_tag_present(skb)) { |
018c1dda EG |
350 | key->eth.vlan.tci = htons(skb->vlan_tci); |
351 | key->eth.vlan.tpid = skb->vlan_proto; | |
352 | } else { | |
353 | /* Parse outer vlan tag in the non-accelerated case. */ | |
354 | res = parse_vlan_tag(skb, &key->eth.vlan); | |
355 | if (res <= 0) | |
356 | return res; | |
357 | } | |
358 | ||
359 | /* Parse inner vlan tag. */ | |
360 | res = parse_vlan_tag(skb, &key->eth.cvlan); | |
361 | if (res <= 0) | |
362 | return res; | |
ccb1352e JG |
363 | |
364 | return 0; | |
365 | } | |
366 | ||
367 | static __be16 parse_ethertype(struct sk_buff *skb) | |
368 | { | |
369 | struct llc_snap_hdr { | |
370 | u8 dsap; /* Always 0xAA */ | |
371 | u8 ssap; /* Always 0xAA */ | |
372 | u8 ctrl; | |
373 | u8 oui[3]; | |
374 | __be16 ethertype; | |
375 | }; | |
376 | struct llc_snap_hdr *llc; | |
377 | __be16 proto; | |
378 | ||
379 | proto = *(__be16 *) skb->data; | |
380 | __skb_pull(skb, sizeof(__be16)); | |
381 | ||
6713fc9b | 382 | if (eth_proto_is_802_3(proto)) |
ccb1352e JG |
383 | return proto; |
384 | ||
385 | if (skb->len < sizeof(struct llc_snap_hdr)) | |
386 | return htons(ETH_P_802_2); | |
387 | ||
388 | if (unlikely(!pskb_may_pull(skb, sizeof(struct llc_snap_hdr)))) | |
389 | return htons(0); | |
390 | ||
391 | llc = (struct llc_snap_hdr *) skb->data; | |
392 | if (llc->dsap != LLC_SAP_SNAP || | |
393 | llc->ssap != LLC_SAP_SNAP || | |
394 | (llc->oui[0] | llc->oui[1] | llc->oui[2]) != 0) | |
395 | return htons(ETH_P_802_2); | |
396 | ||
397 | __skb_pull(skb, sizeof(struct llc_snap_hdr)); | |
17b682a0 | 398 | |
6713fc9b | 399 | if (eth_proto_is_802_3(llc->ethertype)) |
17b682a0 RL |
400 | return llc->ethertype; |
401 | ||
402 | return htons(ETH_P_802_2); | |
ccb1352e JG |
403 | } |
404 | ||
405 | static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, | |
03f0d916 | 406 | int nh_len) |
ccb1352e JG |
407 | { |
408 | struct icmp6hdr *icmp = icmp6_hdr(skb); | |
ccb1352e JG |
409 | |
410 | /* The ICMPv6 type and code fields use the 16-bit transport port | |
411 | * fields, so we need to store them in 16-bit network byte order. | |
412 | */ | |
1139e241 JR |
413 | key->tp.src = htons(icmp->icmp6_type); |
414 | key->tp.dst = htons(icmp->icmp6_code); | |
25ef1328 | 415 | memset(&key->ipv6.nd, 0, sizeof(key->ipv6.nd)); |
ccb1352e JG |
416 | |
417 | if (icmp->icmp6_code == 0 && | |
418 | (icmp->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION || | |
419 | icmp->icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT)) { | |
420 | int icmp_len = skb->len - skb_transport_offset(skb); | |
421 | struct nd_msg *nd; | |
422 | int offset; | |
423 | ||
ccb1352e JG |
424 | /* In order to process neighbor discovery options, we need the |
425 | * entire packet. | |
426 | */ | |
427 | if (unlikely(icmp_len < sizeof(*nd))) | |
03f0d916 AZ |
428 | return 0; |
429 | ||
430 | if (unlikely(skb_linearize(skb))) | |
431 | return -ENOMEM; | |
ccb1352e JG |
432 | |
433 | nd = (struct nd_msg *)skb_transport_header(skb); | |
434 | key->ipv6.nd.target = nd->target; | |
ccb1352e JG |
435 | |
436 | icmp_len -= sizeof(*nd); | |
437 | offset = 0; | |
438 | while (icmp_len >= 8) { | |
439 | struct nd_opt_hdr *nd_opt = | |
440 | (struct nd_opt_hdr *)(nd->opt + offset); | |
441 | int opt_len = nd_opt->nd_opt_len * 8; | |
442 | ||
443 | if (unlikely(!opt_len || opt_len > icmp_len)) | |
03f0d916 | 444 | return 0; |
ccb1352e JG |
445 | |
446 | /* Store the link layer address if the appropriate | |
447 | * option is provided. It is considered an error if | |
448 | * the same link layer option is specified twice. | |
449 | */ | |
450 | if (nd_opt->nd_opt_type == ND_OPT_SOURCE_LL_ADDR | |
451 | && opt_len == 8) { | |
452 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.sll))) | |
453 | goto invalid; | |
8c63ff09 JP |
454 | ether_addr_copy(key->ipv6.nd.sll, |
455 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
456 | } else if (nd_opt->nd_opt_type == ND_OPT_TARGET_LL_ADDR |
457 | && opt_len == 8) { | |
458 | if (unlikely(!is_zero_ether_addr(key->ipv6.nd.tll))) | |
459 | goto invalid; | |
8c63ff09 JP |
460 | ether_addr_copy(key->ipv6.nd.tll, |
461 | &nd->opt[offset+sizeof(*nd_opt)]); | |
ccb1352e JG |
462 | } |
463 | ||
464 | icmp_len -= opt_len; | |
465 | offset += opt_len; | |
466 | } | |
467 | } | |
468 | ||
03f0d916 | 469 | return 0; |
ccb1352e JG |
470 | |
471 | invalid: | |
472 | memset(&key->ipv6.nd.target, 0, sizeof(key->ipv6.nd.target)); | |
473 | memset(key->ipv6.nd.sll, 0, sizeof(key->ipv6.nd.sll)); | |
474 | memset(key->ipv6.nd.tll, 0, sizeof(key->ipv6.nd.tll)); | |
475 | ||
03f0d916 | 476 | return 0; |
ccb1352e JG |
477 | } |
478 | ||
479 | /** | |
83c8df26 | 480 | * key_extract - extracts a flow key from an Ethernet frame. |
ccb1352e JG |
481 | * @skb: sk_buff that contains the frame, with skb->data pointing to the |
482 | * Ethernet header | |
ccb1352e | 483 | * @key: output flow key |
ccb1352e JG |
484 | * |
485 | * The caller must ensure that skb->len >= ETH_HLEN. | |
486 | * | |
487 | * Returns 0 if successful, otherwise a negative errno value. | |
488 | * | |
5108bbad | 489 | * Initializes @skb header fields as follows: |
ccb1352e | 490 | * |
5108bbad | 491 | * - skb->mac_header: the L2 header. |
ccb1352e | 492 | * |
5108bbad JB |
493 | * - skb->network_header: just past the L2 header, or just past the |
494 | * VLAN header, to the first byte of the L2 payload. | |
ccb1352e | 495 | * |
34d94f21 | 496 | * - skb->transport_header: If key->eth.type is ETH_P_IP or ETH_P_IPV6 |
ccb1352e JG |
497 | * on output, then just past the IP header, if one is present and |
498 | * of a correct length, otherwise the same as skb->network_header. | |
34d94f21 | 499 | * For other key->eth.type values it is left untouched. |
5108bbad JB |
500 | * |
501 | * - skb->protocol: the type of the data starting at skb->network_header. | |
502 | * Equals to key->eth.type. | |
ccb1352e | 503 | */ |
83c8df26 | 504 | static int key_extract(struct sk_buff *skb, struct sw_flow_key *key) |
ccb1352e | 505 | { |
03f0d916 | 506 | int error; |
ccb1352e JG |
507 | struct ethhdr *eth; |
508 | ||
f5796684 JG |
509 | /* Flags are always used as part of stats */ |
510 | key->tp.flags = 0; | |
511 | ||
ccb1352e JG |
512 | skb_reset_mac_header(skb); |
513 | ||
5108bbad JB |
514 | /* Link layer. */ |
515 | clear_vlan(key); | |
516 | if (key->mac_proto == MAC_PROTO_NONE) { | |
517 | if (unlikely(eth_type_vlan(skb->protocol))) | |
518 | return -EINVAL; | |
ccb1352e | 519 | |
5108bbad JB |
520 | skb_reset_network_header(skb); |
521 | } else { | |
522 | eth = eth_hdr(skb); | |
523 | ether_addr_copy(key->eth.src, eth->h_source); | |
524 | ether_addr_copy(key->eth.dst, eth->h_dest); | |
ccb1352e | 525 | |
5108bbad JB |
526 | __skb_pull(skb, 2 * ETH_ALEN); |
527 | /* We are going to push all headers that we pull, so no need to | |
528 | * update skb->csum here. | |
529 | */ | |
ccb1352e | 530 | |
5108bbad JB |
531 | if (unlikely(parse_vlan(skb, key))) |
532 | return -ENOMEM; | |
533 | ||
534 | skb->protocol = parse_ethertype(skb); | |
535 | if (unlikely(skb->protocol == htons(0))) | |
536 | return -ENOMEM; | |
ccb1352e | 537 | |
5108bbad JB |
538 | skb_reset_network_header(skb); |
539 | __skb_push(skb, skb->data - skb_mac_header(skb)); | |
540 | } | |
25cd9ba0 | 541 | skb_reset_mac_len(skb); |
5108bbad | 542 | key->eth.type = skb->protocol; |
ccb1352e JG |
543 | |
544 | /* Network layer. */ | |
545 | if (key->eth.type == htons(ETH_P_IP)) { | |
546 | struct iphdr *nh; | |
547 | __be16 offset; | |
548 | ||
ccb1352e JG |
549 | error = check_iphdr(skb); |
550 | if (unlikely(error)) { | |
07148121 JG |
551 | memset(&key->ip, 0, sizeof(key->ip)); |
552 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e JG |
553 | if (error == -EINVAL) { |
554 | skb->transport_header = skb->network_header; | |
555 | error = 0; | |
556 | } | |
03f0d916 | 557 | return error; |
ccb1352e JG |
558 | } |
559 | ||
560 | nh = ip_hdr(skb); | |
561 | key->ipv4.addr.src = nh->saddr; | |
562 | key->ipv4.addr.dst = nh->daddr; | |
563 | ||
564 | key->ip.proto = nh->protocol; | |
565 | key->ip.tos = nh->tos; | |
566 | key->ip.ttl = nh->ttl; | |
567 | ||
568 | offset = nh->frag_off & htons(IP_OFFSET); | |
569 | if (offset) { | |
570 | key->ip.frag = OVS_FRAG_TYPE_LATER; | |
03f0d916 | 571 | return 0; |
ccb1352e JG |
572 | } |
573 | if (nh->frag_off & htons(IP_MF) || | |
07148121 | 574 | skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
ccb1352e | 575 | key->ip.frag = OVS_FRAG_TYPE_FIRST; |
07148121 JG |
576 | else |
577 | key->ip.frag = OVS_FRAG_TYPE_NONE; | |
ccb1352e JG |
578 | |
579 | /* Transport layer. */ | |
580 | if (key->ip.proto == IPPROTO_TCP) { | |
ccb1352e JG |
581 | if (tcphdr_ok(skb)) { |
582 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
583 | key->tp.src = tcp->source; |
584 | key->tp.dst = tcp->dest; | |
585 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
586 | } else { |
587 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 588 | } |
07148121 | 589 | |
ccb1352e | 590 | } else if (key->ip.proto == IPPROTO_UDP) { |
ccb1352e JG |
591 | if (udphdr_ok(skb)) { |
592 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
593 | key->tp.src = udp->source; |
594 | key->tp.dst = udp->dest; | |
07148121 JG |
595 | } else { |
596 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 597 | } |
a175a723 JS |
598 | } else if (key->ip.proto == IPPROTO_SCTP) { |
599 | if (sctphdr_ok(skb)) { | |
600 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
601 | key->tp.src = sctp->source; |
602 | key->tp.dst = sctp->dest; | |
07148121 JG |
603 | } else { |
604 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 605 | } |
ccb1352e | 606 | } else if (key->ip.proto == IPPROTO_ICMP) { |
ccb1352e JG |
607 | if (icmphdr_ok(skb)) { |
608 | struct icmphdr *icmp = icmp_hdr(skb); | |
609 | /* The ICMP type and code fields use the 16-bit | |
610 | * transport port fields, so we need to store | |
611 | * them in 16-bit network byte order. */ | |
1139e241 JR |
612 | key->tp.src = htons(icmp->type); |
613 | key->tp.dst = htons(icmp->code); | |
07148121 JG |
614 | } else { |
615 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
616 | } |
617 | } | |
618 | ||
07148121 JG |
619 | } else if (key->eth.type == htons(ETH_P_ARP) || |
620 | key->eth.type == htons(ETH_P_RARP)) { | |
ccb1352e | 621 | struct arp_eth_header *arp; |
389f4894 | 622 | bool arp_available = arphdr_ok(skb); |
ccb1352e JG |
623 | |
624 | arp = (struct arp_eth_header *)skb_network_header(skb); | |
625 | ||
389f4894 | 626 | if (arp_available && |
07148121 JG |
627 | arp->ar_hrd == htons(ARPHRD_ETHER) && |
628 | arp->ar_pro == htons(ETH_P_IP) && | |
629 | arp->ar_hln == ETH_ALEN && | |
630 | arp->ar_pln == 4) { | |
ccb1352e JG |
631 | |
632 | /* We only match on the lower 8 bits of the opcode. */ | |
633 | if (ntohs(arp->ar_op) <= 0xff) | |
634 | key->ip.proto = ntohs(arp->ar_op); | |
07148121 JG |
635 | else |
636 | key->ip.proto = 0; | |
637 | ||
d04d3829 MM |
638 | memcpy(&key->ipv4.addr.src, arp->ar_sip, sizeof(key->ipv4.addr.src)); |
639 | memcpy(&key->ipv4.addr.dst, arp->ar_tip, sizeof(key->ipv4.addr.dst)); | |
8c63ff09 JP |
640 | ether_addr_copy(key->ipv4.arp.sha, arp->ar_sha); |
641 | ether_addr_copy(key->ipv4.arp.tha, arp->ar_tha); | |
07148121 JG |
642 | } else { |
643 | memset(&key->ip, 0, sizeof(key->ip)); | |
644 | memset(&key->ipv4, 0, sizeof(key->ipv4)); | |
ccb1352e | 645 | } |
25cd9ba0 SH |
646 | } else if (eth_p_mpls(key->eth.type)) { |
647 | size_t stack_len = MPLS_HLEN; | |
648 | ||
f7d49bce | 649 | skb_set_inner_network_header(skb, skb->mac_len); |
25cd9ba0 SH |
650 | while (1) { |
651 | __be32 lse; | |
652 | ||
653 | error = check_header(skb, skb->mac_len + stack_len); | |
654 | if (unlikely(error)) | |
655 | return 0; | |
656 | ||
f7d49bce | 657 | memcpy(&lse, skb_inner_network_header(skb), MPLS_HLEN); |
25cd9ba0 SH |
658 | |
659 | if (stack_len == MPLS_HLEN) | |
660 | memcpy(&key->mpls.top_lse, &lse, MPLS_HLEN); | |
661 | ||
f7d49bce | 662 | skb_set_inner_network_header(skb, skb->mac_len + stack_len); |
25cd9ba0 SH |
663 | if (lse & htonl(MPLS_LS_S_MASK)) |
664 | break; | |
665 | ||
666 | stack_len += MPLS_HLEN; | |
667 | } | |
ccb1352e JG |
668 | } else if (key->eth.type == htons(ETH_P_IPV6)) { |
669 | int nh_len; /* IPv6 Header + Extensions */ | |
670 | ||
03f0d916 | 671 | nh_len = parse_ipv6hdr(skb, key); |
ccb1352e | 672 | if (unlikely(nh_len < 0)) { |
c30da497 SH |
673 | switch (nh_len) { |
674 | case -EINVAL: | |
675 | memset(&key->ip, 0, sizeof(key->ip)); | |
676 | memset(&key->ipv6.addr, 0, sizeof(key->ipv6.addr)); | |
677 | /* fall-through */ | |
678 | case -EPROTO: | |
ccb1352e | 679 | skb->transport_header = skb->network_header; |
03f0d916 | 680 | error = 0; |
c30da497 SH |
681 | break; |
682 | default: | |
ccb1352e | 683 | error = nh_len; |
03f0d916 AZ |
684 | } |
685 | return error; | |
ccb1352e JG |
686 | } |
687 | ||
688 | if (key->ip.frag == OVS_FRAG_TYPE_LATER) | |
03f0d916 | 689 | return 0; |
ccb1352e JG |
690 | if (skb_shinfo(skb)->gso_type & SKB_GSO_UDP) |
691 | key->ip.frag = OVS_FRAG_TYPE_FIRST; | |
692 | ||
693 | /* Transport layer. */ | |
694 | if (key->ip.proto == NEXTHDR_TCP) { | |
ccb1352e JG |
695 | if (tcphdr_ok(skb)) { |
696 | struct tcphdr *tcp = tcp_hdr(skb); | |
1139e241 JR |
697 | key->tp.src = tcp->source; |
698 | key->tp.dst = tcp->dest; | |
699 | key->tp.flags = TCP_FLAGS_BE16(tcp); | |
07148121 JG |
700 | } else { |
701 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
702 | } |
703 | } else if (key->ip.proto == NEXTHDR_UDP) { | |
ccb1352e JG |
704 | if (udphdr_ok(skb)) { |
705 | struct udphdr *udp = udp_hdr(skb); | |
1139e241 JR |
706 | key->tp.src = udp->source; |
707 | key->tp.dst = udp->dest; | |
07148121 JG |
708 | } else { |
709 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e | 710 | } |
a175a723 JS |
711 | } else if (key->ip.proto == NEXTHDR_SCTP) { |
712 | if (sctphdr_ok(skb)) { | |
713 | struct sctphdr *sctp = sctp_hdr(skb); | |
1139e241 JR |
714 | key->tp.src = sctp->source; |
715 | key->tp.dst = sctp->dest; | |
07148121 JG |
716 | } else { |
717 | memset(&key->tp, 0, sizeof(key->tp)); | |
a175a723 | 718 | } |
ccb1352e | 719 | } else if (key->ip.proto == NEXTHDR_ICMP) { |
ccb1352e | 720 | if (icmp6hdr_ok(skb)) { |
03f0d916 AZ |
721 | error = parse_icmpv6(skb, key, nh_len); |
722 | if (error) | |
723 | return error; | |
07148121 JG |
724 | } else { |
725 | memset(&key->tp, 0, sizeof(key->tp)); | |
ccb1352e JG |
726 | } |
727 | } | |
728 | } | |
03f0d916 | 729 | return 0; |
ccb1352e | 730 | } |
83c8df26 | 731 | |
971427f3 AZ |
732 | int ovs_flow_key_update(struct sk_buff *skb, struct sw_flow_key *key) |
733 | { | |
734 | return key_extract(skb, key); | |
735 | } | |
736 | ||
5108bbad JB |
737 | static int key_extract_mac_proto(struct sk_buff *skb) |
738 | { | |
739 | switch (skb->dev->type) { | |
740 | case ARPHRD_ETHER: | |
741 | return MAC_PROTO_ETHERNET; | |
742 | case ARPHRD_NONE: | |
743 | if (skb->protocol == htons(ETH_P_TEB)) | |
744 | return MAC_PROTO_ETHERNET; | |
745 | return MAC_PROTO_NONE; | |
746 | } | |
747 | WARN_ON_ONCE(1); | |
748 | return -EINVAL; | |
749 | } | |
750 | ||
1d8fff90 | 751 | int ovs_flow_key_extract(const struct ip_tunnel_info *tun_info, |
8c8b1b83 | 752 | struct sk_buff *skb, struct sw_flow_key *key) |
83c8df26 | 753 | { |
5108bbad JB |
754 | int res; |
755 | ||
83c8df26 | 756 | /* Extract metadata from packet. */ |
f5796684 | 757 | if (tun_info) { |
00a93bab | 758 | key->tun_proto = ip_tunnel_info_af(tun_info); |
1d8fff90 | 759 | memcpy(&key->tun_key, &tun_info->key, sizeof(key->tun_key)); |
f5796684 | 760 | |
4c222798 | 761 | if (tun_info->options_len) { |
f5796684 JG |
762 | BUILD_BUG_ON((1 << (sizeof(tun_info->options_len) * |
763 | 8)) - 1 | |
764 | > sizeof(key->tun_opts)); | |
4c222798 PS |
765 | |
766 | ip_tunnel_info_opts_get(TUN_METADATA_OPTS(key, tun_info->options_len), | |
767 | tun_info); | |
f5796684 JG |
768 | key->tun_opts_len = tun_info->options_len; |
769 | } else { | |
770 | key->tun_opts_len = 0; | |
771 | } | |
772 | } else { | |
00a93bab | 773 | key->tun_proto = 0; |
f5796684 | 774 | key->tun_opts_len = 0; |
07148121 | 775 | memset(&key->tun_key, 0, sizeof(key->tun_key)); |
f5796684 | 776 | } |
83c8df26 PS |
777 | |
778 | key->phy.priority = skb->priority; | |
779 | key->phy.in_port = OVS_CB(skb)->input_vport->port_no; | |
780 | key->phy.skb_mark = skb->mark; | |
7f8a436e | 781 | ovs_ct_fill_key(skb, key); |
07148121 | 782 | key->ovs_flow_hash = 0; |
5108bbad JB |
783 | res = key_extract_mac_proto(skb); |
784 | if (res < 0) | |
785 | return res; | |
786 | key->mac_proto = res; | |
07148121 JG |
787 | key->recirc_id = 0; |
788 | ||
83c8df26 PS |
789 | return key_extract(skb, key); |
790 | } | |
791 | ||
c2ac6673 | 792 | int ovs_flow_key_extract_userspace(struct net *net, const struct nlattr *attr, |
83c8df26 | 793 | struct sk_buff *skb, |
05da5898 | 794 | struct sw_flow_key *key, bool log) |
83c8df26 PS |
795 | { |
796 | int err; | |
797 | ||
83c8df26 | 798 | /* Extract metadata from netlink attributes. */ |
c2ac6673 | 799 | err = ovs_nla_get_flow_metadata(net, attr, key, log); |
83c8df26 PS |
800 | if (err) |
801 | return err; | |
802 | ||
5108bbad JB |
803 | if (ovs_key_mac_proto(key) == MAC_PROTO_NONE) { |
804 | /* key_extract assumes that skb->protocol is set-up for | |
805 | * layer 3 packets which is the case for other callers, | |
806 | * in particular packets recieved from the network stack. | |
807 | * Here the correct value can be set from the metadata | |
808 | * extracted above. | |
809 | */ | |
810 | skb->protocol = key->eth.type; | |
811 | } else { | |
812 | struct ethhdr *eth; | |
813 | ||
814 | skb_reset_mac_header(skb); | |
815 | eth = eth_hdr(skb); | |
816 | ||
817 | /* Normally, setting the skb 'protocol' field would be | |
818 | * handled by a call to eth_type_trans(), but it assumes | |
819 | * there's a sending device, which we may not have. | |
820 | */ | |
821 | if (eth_proto_is_802_3(eth->h_proto)) | |
822 | skb->protocol = eth->h_proto; | |
823 | else | |
824 | skb->protocol = htons(ETH_P_802_2); | |
825 | } | |
826 | ||
83c8df26 PS |
827 | return key_extract(skb, key); |
828 | } |