xref: /openbmc/linux/net/core/flow_dissector.c (revision 0edbfea5)
1 #include <linux/kernel.h>
2 #include <linux/skbuff.h>
3 #include <linux/export.h>
4 #include <linux/ip.h>
5 #include <linux/ipv6.h>
6 #include <linux/if_vlan.h>
7 #include <net/ip.h>
8 #include <net/ipv6.h>
9 #include <linux/igmp.h>
10 #include <linux/icmp.h>
11 #include <linux/sctp.h>
12 #include <linux/dccp.h>
13 #include <linux/if_tunnel.h>
14 #include <linux/if_pppox.h>
15 #include <linux/ppp_defs.h>
16 #include <linux/stddef.h>
17 #include <linux/if_ether.h>
18 #include <linux/mpls.h>
19 #include <net/flow_dissector.h>
20 #include <scsi/fc/fc_fcoe.h>
21 
22 static void dissector_set_key(struct flow_dissector *flow_dissector,
23 			      enum flow_dissector_key_id key_id)
24 {
25 	flow_dissector->used_keys |= (1 << key_id);
26 }
27 
28 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
29 			     const struct flow_dissector_key *key,
30 			     unsigned int key_count)
31 {
32 	unsigned int i;
33 
34 	memset(flow_dissector, 0, sizeof(*flow_dissector));
35 
36 	for (i = 0; i < key_count; i++, key++) {
37 		/* User should make sure that every key target offset is withing
38 		 * boundaries of unsigned short.
39 		 */
40 		BUG_ON(key->offset > USHRT_MAX);
41 		BUG_ON(dissector_uses_key(flow_dissector,
42 					  key->key_id));
43 
44 		dissector_set_key(flow_dissector, key->key_id);
45 		flow_dissector->offset[key->key_id] = key->offset;
46 	}
47 
48 	/* Ensure that the dissector always includes control and basic key.
49 	 * That way we are able to avoid handling lack of these in fast path.
50 	 */
51 	BUG_ON(!dissector_uses_key(flow_dissector,
52 				   FLOW_DISSECTOR_KEY_CONTROL));
53 	BUG_ON(!dissector_uses_key(flow_dissector,
54 				   FLOW_DISSECTOR_KEY_BASIC));
55 }
56 EXPORT_SYMBOL(skb_flow_dissector_init);
57 
58 /**
59  * __skb_flow_get_ports - extract the upper layer ports and return them
60  * @skb: sk_buff to extract the ports from
61  * @thoff: transport header offset
62  * @ip_proto: protocol for which to get port offset
63  * @data: raw buffer pointer to the packet, if NULL use skb->data
64  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
65  *
66  * The function will try to retrieve the ports at offset thoff + poff where poff
67  * is the protocol port offset returned from proto_ports_offset
68  */
69 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
70 			    void *data, int hlen)
71 {
72 	int poff = proto_ports_offset(ip_proto);
73 
74 	if (!data) {
75 		data = skb->data;
76 		hlen = skb_headlen(skb);
77 	}
78 
79 	if (poff >= 0) {
80 		__be32 *ports, _ports;
81 
82 		ports = __skb_header_pointer(skb, thoff + poff,
83 					     sizeof(_ports), data, hlen, &_ports);
84 		if (ports)
85 			return *ports;
86 	}
87 
88 	return 0;
89 }
90 EXPORT_SYMBOL(__skb_flow_get_ports);
91 
92 /**
93  * __skb_flow_dissect - extract the flow_keys struct and return it
94  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
95  * @flow_dissector: list of keys to dissect
96  * @target_container: target structure to put dissected values into
97  * @data: raw buffer pointer to the packet, if NULL use skb->data
98  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
99  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
100  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
101  *
102  * The function will try to retrieve individual keys into target specified
103  * by flow_dissector from either the skbuff or a raw buffer specified by the
104  * rest parameters.
105  *
106  * Caller must take care of zeroing target container memory.
107  */
108 bool __skb_flow_dissect(const struct sk_buff *skb,
109 			struct flow_dissector *flow_dissector,
110 			void *target_container,
111 			void *data, __be16 proto, int nhoff, int hlen,
112 			unsigned int flags)
113 {
114 	struct flow_dissector_key_control *key_control;
115 	struct flow_dissector_key_basic *key_basic;
116 	struct flow_dissector_key_addrs *key_addrs;
117 	struct flow_dissector_key_ports *key_ports;
118 	struct flow_dissector_key_tags *key_tags;
119 	struct flow_dissector_key_keyid *key_keyid;
120 	u8 ip_proto = 0;
121 	bool ret = false;
122 
123 	if (!data) {
124 		data = skb->data;
125 		proto = skb->protocol;
126 		nhoff = skb_network_offset(skb);
127 		hlen = skb_headlen(skb);
128 	}
129 
130 	/* It is ensured by skb_flow_dissector_init() that control key will
131 	 * be always present.
132 	 */
133 	key_control = skb_flow_dissector_target(flow_dissector,
134 						FLOW_DISSECTOR_KEY_CONTROL,
135 						target_container);
136 
137 	/* It is ensured by skb_flow_dissector_init() that basic key will
138 	 * be always present.
139 	 */
140 	key_basic = skb_flow_dissector_target(flow_dissector,
141 					      FLOW_DISSECTOR_KEY_BASIC,
142 					      target_container);
143 
144 	if (dissector_uses_key(flow_dissector,
145 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
146 		struct ethhdr *eth = eth_hdr(skb);
147 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
148 
149 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
150 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
151 							  target_container);
152 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
153 	}
154 
155 again:
156 	switch (proto) {
157 	case htons(ETH_P_IP): {
158 		const struct iphdr *iph;
159 		struct iphdr _iph;
160 ip:
161 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
162 		if (!iph || iph->ihl < 5)
163 			goto out_bad;
164 		nhoff += iph->ihl * 4;
165 
166 		ip_proto = iph->protocol;
167 
168 		if (dissector_uses_key(flow_dissector,
169 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
170 			key_addrs = skb_flow_dissector_target(flow_dissector,
171 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
172 							      target_container);
173 
174 			memcpy(&key_addrs->v4addrs, &iph->saddr,
175 			       sizeof(key_addrs->v4addrs));
176 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
177 		}
178 
179 		if (ip_is_fragment(iph)) {
180 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
181 
182 			if (iph->frag_off & htons(IP_OFFSET)) {
183 				goto out_good;
184 			} else {
185 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
186 				if (!(flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG))
187 					goto out_good;
188 			}
189 		}
190 
191 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
192 			goto out_good;
193 
194 		break;
195 	}
196 	case htons(ETH_P_IPV6): {
197 		const struct ipv6hdr *iph;
198 		struct ipv6hdr _iph;
199 
200 ipv6:
201 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
202 		if (!iph)
203 			goto out_bad;
204 
205 		ip_proto = iph->nexthdr;
206 		nhoff += sizeof(struct ipv6hdr);
207 
208 		if (dissector_uses_key(flow_dissector,
209 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
210 			key_addrs = skb_flow_dissector_target(flow_dissector,
211 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
212 							      target_container);
213 
214 			memcpy(&key_addrs->v6addrs, &iph->saddr,
215 			       sizeof(key_addrs->v6addrs));
216 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
217 		}
218 
219 		if ((dissector_uses_key(flow_dissector,
220 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
221 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
222 		    ip6_flowlabel(iph)) {
223 			__be32 flow_label = ip6_flowlabel(iph);
224 
225 			if (dissector_uses_key(flow_dissector,
226 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
227 				key_tags = skb_flow_dissector_target(flow_dissector,
228 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
229 								     target_container);
230 				key_tags->flow_label = ntohl(flow_label);
231 			}
232 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)
233 				goto out_good;
234 		}
235 
236 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
237 			goto out_good;
238 
239 		break;
240 	}
241 	case htons(ETH_P_8021AD):
242 	case htons(ETH_P_8021Q): {
243 		const struct vlan_hdr *vlan;
244 		struct vlan_hdr _vlan;
245 
246 		vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan);
247 		if (!vlan)
248 			goto out_bad;
249 
250 		if (dissector_uses_key(flow_dissector,
251 				       FLOW_DISSECTOR_KEY_VLANID)) {
252 			key_tags = skb_flow_dissector_target(flow_dissector,
253 							     FLOW_DISSECTOR_KEY_VLANID,
254 							     target_container);
255 
256 			key_tags->vlan_id = skb_vlan_tag_get_id(skb);
257 		}
258 
259 		proto = vlan->h_vlan_encapsulated_proto;
260 		nhoff += sizeof(*vlan);
261 		goto again;
262 	}
263 	case htons(ETH_P_PPP_SES): {
264 		struct {
265 			struct pppoe_hdr hdr;
266 			__be16 proto;
267 		} *hdr, _hdr;
268 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
269 		if (!hdr)
270 			goto out_bad;
271 		proto = hdr->proto;
272 		nhoff += PPPOE_SES_HLEN;
273 		switch (proto) {
274 		case htons(PPP_IP):
275 			goto ip;
276 		case htons(PPP_IPV6):
277 			goto ipv6;
278 		default:
279 			goto out_bad;
280 		}
281 	}
282 	case htons(ETH_P_TIPC): {
283 		struct {
284 			__be32 pre[3];
285 			__be32 srcnode;
286 		} *hdr, _hdr;
287 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
288 		if (!hdr)
289 			goto out_bad;
290 
291 		if (dissector_uses_key(flow_dissector,
292 				       FLOW_DISSECTOR_KEY_TIPC_ADDRS)) {
293 			key_addrs = skb_flow_dissector_target(flow_dissector,
294 							      FLOW_DISSECTOR_KEY_TIPC_ADDRS,
295 							      target_container);
296 			key_addrs->tipcaddrs.srcnode = hdr->srcnode;
297 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS;
298 		}
299 		goto out_good;
300 	}
301 
302 	case htons(ETH_P_MPLS_UC):
303 	case htons(ETH_P_MPLS_MC): {
304 		struct mpls_label *hdr, _hdr[2];
305 mpls:
306 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
307 					   hlen, &_hdr);
308 		if (!hdr)
309 			goto out_bad;
310 
311 		if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >>
312 		     MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) {
313 			if (dissector_uses_key(flow_dissector,
314 					       FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
315 				key_keyid = skb_flow_dissector_target(flow_dissector,
316 								      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
317 								      target_container);
318 				key_keyid->keyid = hdr[1].entry &
319 					htonl(MPLS_LS_LABEL_MASK);
320 			}
321 
322 			goto out_good;
323 		}
324 
325 		goto out_good;
326 	}
327 
328 	case htons(ETH_P_FCOE):
329 		if ((hlen - nhoff) < FCOE_HEADER_LEN)
330 			goto out_bad;
331 
332 		nhoff += FCOE_HEADER_LEN;
333 		goto out_good;
334 	default:
335 		goto out_bad;
336 	}
337 
338 ip_proto_again:
339 	switch (ip_proto) {
340 	case IPPROTO_GRE: {
341 		struct gre_hdr {
342 			__be16 flags;
343 			__be16 proto;
344 		} *hdr, _hdr;
345 
346 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
347 		if (!hdr)
348 			goto out_bad;
349 		/*
350 		 * Only look inside GRE if version zero and no
351 		 * routing
352 		 */
353 		if (hdr->flags & (GRE_VERSION | GRE_ROUTING))
354 			break;
355 
356 		proto = hdr->proto;
357 		nhoff += 4;
358 		if (hdr->flags & GRE_CSUM)
359 			nhoff += 4;
360 		if (hdr->flags & GRE_KEY) {
361 			const __be32 *keyid;
362 			__be32 _keyid;
363 
364 			keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid),
365 						     data, hlen, &_keyid);
366 
367 			if (!keyid)
368 				goto out_bad;
369 
370 			if (dissector_uses_key(flow_dissector,
371 					       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
372 				key_keyid = skb_flow_dissector_target(flow_dissector,
373 								      FLOW_DISSECTOR_KEY_GRE_KEYID,
374 								      target_container);
375 				key_keyid->keyid = *keyid;
376 			}
377 			nhoff += 4;
378 		}
379 		if (hdr->flags & GRE_SEQ)
380 			nhoff += 4;
381 		if (proto == htons(ETH_P_TEB)) {
382 			const struct ethhdr *eth;
383 			struct ethhdr _eth;
384 
385 			eth = __skb_header_pointer(skb, nhoff,
386 						   sizeof(_eth),
387 						   data, hlen, &_eth);
388 			if (!eth)
389 				goto out_bad;
390 			proto = eth->h_proto;
391 			nhoff += sizeof(*eth);
392 
393 			/* Cap headers that we access via pointers at the
394 			 * end of the Ethernet header as our maximum alignment
395 			 * at that point is only 2 bytes.
396 			 */
397 			if (NET_IP_ALIGN)
398 				hlen = nhoff;
399 		}
400 
401 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
402 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
403 			goto out_good;
404 
405 		goto again;
406 	}
407 	case NEXTHDR_HOP:
408 	case NEXTHDR_ROUTING:
409 	case NEXTHDR_DEST: {
410 		u8 _opthdr[2], *opthdr;
411 
412 		if (proto != htons(ETH_P_IPV6))
413 			break;
414 
415 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
416 					      data, hlen, &_opthdr);
417 		if (!opthdr)
418 			goto out_bad;
419 
420 		ip_proto = opthdr[0];
421 		nhoff += (opthdr[1] + 1) << 3;
422 
423 		goto ip_proto_again;
424 	}
425 	case NEXTHDR_FRAGMENT: {
426 		struct frag_hdr _fh, *fh;
427 
428 		if (proto != htons(ETH_P_IPV6))
429 			break;
430 
431 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
432 					  data, hlen, &_fh);
433 
434 		if (!fh)
435 			goto out_bad;
436 
437 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
438 
439 		nhoff += sizeof(_fh);
440 		ip_proto = fh->nexthdr;
441 
442 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
443 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
444 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG)
445 				goto ip_proto_again;
446 		}
447 		goto out_good;
448 	}
449 	case IPPROTO_IPIP:
450 		proto = htons(ETH_P_IP);
451 
452 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
453 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
454 			goto out_good;
455 
456 		goto ip;
457 	case IPPROTO_IPV6:
458 		proto = htons(ETH_P_IPV6);
459 
460 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
461 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
462 			goto out_good;
463 
464 		goto ipv6;
465 	case IPPROTO_MPLS:
466 		proto = htons(ETH_P_MPLS_UC);
467 		goto mpls;
468 	default:
469 		break;
470 	}
471 
472 	if (dissector_uses_key(flow_dissector,
473 			       FLOW_DISSECTOR_KEY_PORTS)) {
474 		key_ports = skb_flow_dissector_target(flow_dissector,
475 						      FLOW_DISSECTOR_KEY_PORTS,
476 						      target_container);
477 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
478 							data, hlen);
479 	}
480 
481 out_good:
482 	ret = true;
483 
484 out_bad:
485 	key_basic->n_proto = proto;
486 	key_basic->ip_proto = ip_proto;
487 	key_control->thoff = (u16)nhoff;
488 
489 	return ret;
490 }
491 EXPORT_SYMBOL(__skb_flow_dissect);
492 
493 static u32 hashrnd __read_mostly;
494 static __always_inline void __flow_hash_secret_init(void)
495 {
496 	net_get_random_once(&hashrnd, sizeof(hashrnd));
497 }
498 
499 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
500 					     u32 keyval)
501 {
502 	return jhash2(words, length, keyval);
503 }
504 
505 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
506 {
507 	const void *p = flow;
508 
509 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
510 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
511 }
512 
513 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
514 {
515 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
516 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
517 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
518 		     sizeof(*flow) - sizeof(flow->addrs));
519 
520 	switch (flow->control.addr_type) {
521 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
522 		diff -= sizeof(flow->addrs.v4addrs);
523 		break;
524 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
525 		diff -= sizeof(flow->addrs.v6addrs);
526 		break;
527 	case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
528 		diff -= sizeof(flow->addrs.tipcaddrs);
529 		break;
530 	}
531 	return (sizeof(*flow) - diff) / sizeof(u32);
532 }
533 
534 __be32 flow_get_u32_src(const struct flow_keys *flow)
535 {
536 	switch (flow->control.addr_type) {
537 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
538 		return flow->addrs.v4addrs.src;
539 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
540 		return (__force __be32)ipv6_addr_hash(
541 			&flow->addrs.v6addrs.src);
542 	case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
543 		return flow->addrs.tipcaddrs.srcnode;
544 	default:
545 		return 0;
546 	}
547 }
548 EXPORT_SYMBOL(flow_get_u32_src);
549 
550 __be32 flow_get_u32_dst(const struct flow_keys *flow)
551 {
552 	switch (flow->control.addr_type) {
553 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
554 		return flow->addrs.v4addrs.dst;
555 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
556 		return (__force __be32)ipv6_addr_hash(
557 			&flow->addrs.v6addrs.dst);
558 	default:
559 		return 0;
560 	}
561 }
562 EXPORT_SYMBOL(flow_get_u32_dst);
563 
564 static inline void __flow_hash_consistentify(struct flow_keys *keys)
565 {
566 	int addr_diff, i;
567 
568 	switch (keys->control.addr_type) {
569 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
570 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
571 			    (__force u32)keys->addrs.v4addrs.src;
572 		if ((addr_diff < 0) ||
573 		    (addr_diff == 0 &&
574 		     ((__force u16)keys->ports.dst <
575 		      (__force u16)keys->ports.src))) {
576 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
577 			swap(keys->ports.src, keys->ports.dst);
578 		}
579 		break;
580 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
581 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
582 				   &keys->addrs.v6addrs.src,
583 				   sizeof(keys->addrs.v6addrs.dst));
584 		if ((addr_diff < 0) ||
585 		    (addr_diff == 0 &&
586 		     ((__force u16)keys->ports.dst <
587 		      (__force u16)keys->ports.src))) {
588 			for (i = 0; i < 4; i++)
589 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
590 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
591 			swap(keys->ports.src, keys->ports.dst);
592 		}
593 		break;
594 	}
595 }
596 
597 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
598 {
599 	u32 hash;
600 
601 	__flow_hash_consistentify(keys);
602 
603 	hash = __flow_hash_words(flow_keys_hash_start(keys),
604 				 flow_keys_hash_length(keys), keyval);
605 	if (!hash)
606 		hash = 1;
607 
608 	return hash;
609 }
610 
611 u32 flow_hash_from_keys(struct flow_keys *keys)
612 {
613 	__flow_hash_secret_init();
614 	return __flow_hash_from_keys(keys, hashrnd);
615 }
616 EXPORT_SYMBOL(flow_hash_from_keys);
617 
618 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
619 				  struct flow_keys *keys, u32 keyval)
620 {
621 	skb_flow_dissect_flow_keys(skb, keys,
622 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
623 
624 	return __flow_hash_from_keys(keys, keyval);
625 }
626 
627 struct _flow_keys_digest_data {
628 	__be16	n_proto;
629 	u8	ip_proto;
630 	u8	padding;
631 	__be32	ports;
632 	__be32	src;
633 	__be32	dst;
634 };
635 
636 void make_flow_keys_digest(struct flow_keys_digest *digest,
637 			   const struct flow_keys *flow)
638 {
639 	struct _flow_keys_digest_data *data =
640 	    (struct _flow_keys_digest_data *)digest;
641 
642 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
643 
644 	memset(digest, 0, sizeof(*digest));
645 
646 	data->n_proto = flow->basic.n_proto;
647 	data->ip_proto = flow->basic.ip_proto;
648 	data->ports = flow->ports.ports;
649 	data->src = flow->addrs.v4addrs.src;
650 	data->dst = flow->addrs.v4addrs.dst;
651 }
652 EXPORT_SYMBOL(make_flow_keys_digest);
653 
654 /**
655  * __skb_get_hash: calculate a flow hash
656  * @skb: sk_buff to calculate flow hash from
657  *
658  * This function calculates a flow hash based on src/dst addresses
659  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
660  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
661  * if hash is a canonical 4-tuple hash over transport ports.
662  */
663 void __skb_get_hash(struct sk_buff *skb)
664 {
665 	struct flow_keys keys;
666 
667 	__flow_hash_secret_init();
668 
669 	__skb_set_sw_hash(skb, ___skb_get_hash(skb, &keys, hashrnd),
670 			  flow_keys_have_l4(&keys));
671 }
672 EXPORT_SYMBOL(__skb_get_hash);
673 
674 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
675 {
676 	struct flow_keys keys;
677 
678 	return ___skb_get_hash(skb, &keys, perturb);
679 }
680 EXPORT_SYMBOL(skb_get_hash_perturb);
681 
682 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
683 {
684 	struct flow_keys keys;
685 
686 	memset(&keys, 0, sizeof(keys));
687 
688 	memcpy(&keys.addrs.v6addrs.src, &fl6->saddr,
689 	       sizeof(keys.addrs.v6addrs.src));
690 	memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr,
691 	       sizeof(keys.addrs.v6addrs.dst));
692 	keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
693 	keys.ports.src = fl6->fl6_sport;
694 	keys.ports.dst = fl6->fl6_dport;
695 	keys.keyid.keyid = fl6->fl6_gre_key;
696 	keys.tags.flow_label = (__force u32)fl6->flowlabel;
697 	keys.basic.ip_proto = fl6->flowi6_proto;
698 
699 	__skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
700 			  flow_keys_have_l4(&keys));
701 
702 	return skb->hash;
703 }
704 EXPORT_SYMBOL(__skb_get_hash_flowi6);
705 
706 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
707 {
708 	struct flow_keys keys;
709 
710 	memset(&keys, 0, sizeof(keys));
711 
712 	keys.addrs.v4addrs.src = fl4->saddr;
713 	keys.addrs.v4addrs.dst = fl4->daddr;
714 	keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
715 	keys.ports.src = fl4->fl4_sport;
716 	keys.ports.dst = fl4->fl4_dport;
717 	keys.keyid.keyid = fl4->fl4_gre_key;
718 	keys.basic.ip_proto = fl4->flowi4_proto;
719 
720 	__skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
721 			  flow_keys_have_l4(&keys));
722 
723 	return skb->hash;
724 }
725 EXPORT_SYMBOL(__skb_get_hash_flowi4);
726 
727 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
728 		   const struct flow_keys *keys, int hlen)
729 {
730 	u32 poff = keys->control.thoff;
731 
732 	/* skip L4 headers for fragments after the first */
733 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
734 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
735 		return poff;
736 
737 	switch (keys->basic.ip_proto) {
738 	case IPPROTO_TCP: {
739 		/* access doff as u8 to avoid unaligned access */
740 		const u8 *doff;
741 		u8 _doff;
742 
743 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
744 					    data, hlen, &_doff);
745 		if (!doff)
746 			return poff;
747 
748 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
749 		break;
750 	}
751 	case IPPROTO_UDP:
752 	case IPPROTO_UDPLITE:
753 		poff += sizeof(struct udphdr);
754 		break;
755 	/* For the rest, we do not really care about header
756 	 * extensions at this point for now.
757 	 */
758 	case IPPROTO_ICMP:
759 		poff += sizeof(struct icmphdr);
760 		break;
761 	case IPPROTO_ICMPV6:
762 		poff += sizeof(struct icmp6hdr);
763 		break;
764 	case IPPROTO_IGMP:
765 		poff += sizeof(struct igmphdr);
766 		break;
767 	case IPPROTO_DCCP:
768 		poff += sizeof(struct dccp_hdr);
769 		break;
770 	case IPPROTO_SCTP:
771 		poff += sizeof(struct sctphdr);
772 		break;
773 	}
774 
775 	return poff;
776 }
777 
778 /**
779  * skb_get_poff - get the offset to the payload
780  * @skb: sk_buff to get the payload offset from
781  *
782  * The function will get the offset to the payload as far as it could
783  * be dissected.  The main user is currently BPF, so that we can dynamically
784  * truncate packets without needing to push actual payload to the user
785  * space and can analyze headers only, instead.
786  */
787 u32 skb_get_poff(const struct sk_buff *skb)
788 {
789 	struct flow_keys keys;
790 
791 	if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
792 		return 0;
793 
794 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
795 }
796 
797 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
798 {
799 	memset(keys, 0, sizeof(*keys));
800 
801 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
802 	    sizeof(keys->addrs.v6addrs.src));
803 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
804 	    sizeof(keys->addrs.v6addrs.dst));
805 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
806 	keys->ports.src = fl6->fl6_sport;
807 	keys->ports.dst = fl6->fl6_dport;
808 	keys->keyid.keyid = fl6->fl6_gre_key;
809 	keys->tags.flow_label = (__force u32)fl6->flowlabel;
810 	keys->basic.ip_proto = fl6->flowi6_proto;
811 
812 	return flow_hash_from_keys(keys);
813 }
814 EXPORT_SYMBOL(__get_hash_from_flowi6);
815 
816 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys)
817 {
818 	memset(keys, 0, sizeof(*keys));
819 
820 	keys->addrs.v4addrs.src = fl4->saddr;
821 	keys->addrs.v4addrs.dst = fl4->daddr;
822 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
823 	keys->ports.src = fl4->fl4_sport;
824 	keys->ports.dst = fl4->fl4_dport;
825 	keys->keyid.keyid = fl4->fl4_gre_key;
826 	keys->basic.ip_proto = fl4->flowi4_proto;
827 
828 	return flow_hash_from_keys(keys);
829 }
830 EXPORT_SYMBOL(__get_hash_from_flowi4);
831 
832 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
833 	{
834 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
835 		.offset = offsetof(struct flow_keys, control),
836 	},
837 	{
838 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
839 		.offset = offsetof(struct flow_keys, basic),
840 	},
841 	{
842 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
843 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
844 	},
845 	{
846 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
847 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
848 	},
849 	{
850 		.key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS,
851 		.offset = offsetof(struct flow_keys, addrs.tipcaddrs),
852 	},
853 	{
854 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
855 		.offset = offsetof(struct flow_keys, ports),
856 	},
857 	{
858 		.key_id = FLOW_DISSECTOR_KEY_VLANID,
859 		.offset = offsetof(struct flow_keys, tags),
860 	},
861 	{
862 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
863 		.offset = offsetof(struct flow_keys, tags),
864 	},
865 	{
866 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
867 		.offset = offsetof(struct flow_keys, keyid),
868 	},
869 };
870 
871 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
872 	{
873 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
874 		.offset = offsetof(struct flow_keys, control),
875 	},
876 	{
877 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
878 		.offset = offsetof(struct flow_keys, basic),
879 	},
880 };
881 
882 struct flow_dissector flow_keys_dissector __read_mostly;
883 EXPORT_SYMBOL(flow_keys_dissector);
884 
885 struct flow_dissector flow_keys_buf_dissector __read_mostly;
886 
887 static int __init init_default_flow_dissectors(void)
888 {
889 	skb_flow_dissector_init(&flow_keys_dissector,
890 				flow_keys_dissector_keys,
891 				ARRAY_SIZE(flow_keys_dissector_keys));
892 	skb_flow_dissector_init(&flow_keys_buf_dissector,
893 				flow_keys_buf_dissector_keys,
894 				ARRAY_SIZE(flow_keys_buf_dissector_keys));
895 	return 0;
896 }
897 
898 late_initcall_sync(init_default_flow_dissectors);
899