xref: /openbmc/linux/net/core/flow_dissector.c (revision e639c869)
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/dsa.h>
8 #include <net/dst_metadata.h>
9 #include <net/ip.h>
10 #include <net/ipv6.h>
11 #include <net/gre.h>
12 #include <net/pptp.h>
13 #include <net/tipc.h>
14 #include <linux/igmp.h>
15 #include <linux/icmp.h>
16 #include <linux/sctp.h>
17 #include <linux/dccp.h>
18 #include <linux/if_tunnel.h>
19 #include <linux/if_pppox.h>
20 #include <linux/ppp_defs.h>
21 #include <linux/stddef.h>
22 #include <linux/if_ether.h>
23 #include <linux/mpls.h>
24 #include <linux/tcp.h>
25 #include <net/flow_dissector.h>
26 #include <scsi/fc/fc_fcoe.h>
27 
28 static void dissector_set_key(struct flow_dissector *flow_dissector,
29 			      enum flow_dissector_key_id key_id)
30 {
31 	flow_dissector->used_keys |= (1 << key_id);
32 }
33 
34 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
35 			     const struct flow_dissector_key *key,
36 			     unsigned int key_count)
37 {
38 	unsigned int i;
39 
40 	memset(flow_dissector, 0, sizeof(*flow_dissector));
41 
42 	for (i = 0; i < key_count; i++, key++) {
43 		/* User should make sure that every key target offset is withing
44 		 * boundaries of unsigned short.
45 		 */
46 		BUG_ON(key->offset > USHRT_MAX);
47 		BUG_ON(dissector_uses_key(flow_dissector,
48 					  key->key_id));
49 
50 		dissector_set_key(flow_dissector, key->key_id);
51 		flow_dissector->offset[key->key_id] = key->offset;
52 	}
53 
54 	/* Ensure that the dissector always includes control and basic key.
55 	 * That way we are able to avoid handling lack of these in fast path.
56 	 */
57 	BUG_ON(!dissector_uses_key(flow_dissector,
58 				   FLOW_DISSECTOR_KEY_CONTROL));
59 	BUG_ON(!dissector_uses_key(flow_dissector,
60 				   FLOW_DISSECTOR_KEY_BASIC));
61 }
62 EXPORT_SYMBOL(skb_flow_dissector_init);
63 
64 /**
65  * skb_flow_get_be16 - extract be16 entity
66  * @skb: sk_buff to extract from
67  * @poff: offset to extract at
68  * @data: raw buffer pointer to the packet
69  * @hlen: packet header length
70  *
71  * The function will try to retrieve a be32 entity at
72  * offset poff
73  */
74 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
75 				void *data, int hlen)
76 {
77 	__be16 *u, _u;
78 
79 	u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
80 	if (u)
81 		return *u;
82 
83 	return 0;
84 }
85 
86 /**
87  * __skb_flow_get_ports - extract the upper layer ports and return them
88  * @skb: sk_buff to extract the ports from
89  * @thoff: transport header offset
90  * @ip_proto: protocol for which to get port offset
91  * @data: raw buffer pointer to the packet, if NULL use skb->data
92  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
93  *
94  * The function will try to retrieve the ports at offset thoff + poff where poff
95  * is the protocol port offset returned from proto_ports_offset
96  */
97 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
98 			    void *data, int hlen)
99 {
100 	int poff = proto_ports_offset(ip_proto);
101 
102 	if (!data) {
103 		data = skb->data;
104 		hlen = skb_headlen(skb);
105 	}
106 
107 	if (poff >= 0) {
108 		__be32 *ports, _ports;
109 
110 		ports = __skb_header_pointer(skb, thoff + poff,
111 					     sizeof(_ports), data, hlen, &_ports);
112 		if (ports)
113 			return *ports;
114 	}
115 
116 	return 0;
117 }
118 EXPORT_SYMBOL(__skb_flow_get_ports);
119 
120 static void
121 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
122 				   struct flow_dissector *flow_dissector,
123 				   void *target_container)
124 {
125 	struct flow_dissector_key_control *ctrl;
126 
127 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
128 		return;
129 
130 	ctrl = skb_flow_dissector_target(flow_dissector,
131 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
132 					 target_container);
133 	ctrl->addr_type = type;
134 }
135 
136 static void
137 __skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
138 			       struct flow_dissector *flow_dissector,
139 			       void *target_container)
140 {
141 	struct ip_tunnel_info *info;
142 	struct ip_tunnel_key *key;
143 
144 	/* A quick check to see if there might be something to do. */
145 	if (!dissector_uses_key(flow_dissector,
146 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
147 	    !dissector_uses_key(flow_dissector,
148 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
149 	    !dissector_uses_key(flow_dissector,
150 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
151 	    !dissector_uses_key(flow_dissector,
152 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
153 	    !dissector_uses_key(flow_dissector,
154 				FLOW_DISSECTOR_KEY_ENC_PORTS))
155 		return;
156 
157 	info = skb_tunnel_info(skb);
158 	if (!info)
159 		return;
160 
161 	key = &info->key;
162 
163 	switch (ip_tunnel_info_af(info)) {
164 	case AF_INET:
165 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
166 						   flow_dissector,
167 						   target_container);
168 		if (dissector_uses_key(flow_dissector,
169 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
170 			struct flow_dissector_key_ipv4_addrs *ipv4;
171 
172 			ipv4 = skb_flow_dissector_target(flow_dissector,
173 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
174 							 target_container);
175 			ipv4->src = key->u.ipv4.src;
176 			ipv4->dst = key->u.ipv4.dst;
177 		}
178 		break;
179 	case AF_INET6:
180 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
181 						   flow_dissector,
182 						   target_container);
183 		if (dissector_uses_key(flow_dissector,
184 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
185 			struct flow_dissector_key_ipv6_addrs *ipv6;
186 
187 			ipv6 = skb_flow_dissector_target(flow_dissector,
188 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
189 							 target_container);
190 			ipv6->src = key->u.ipv6.src;
191 			ipv6->dst = key->u.ipv6.dst;
192 		}
193 		break;
194 	}
195 
196 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
197 		struct flow_dissector_key_keyid *keyid;
198 
199 		keyid = skb_flow_dissector_target(flow_dissector,
200 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
201 						  target_container);
202 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
203 	}
204 
205 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
206 		struct flow_dissector_key_ports *tp;
207 
208 		tp = skb_flow_dissector_target(flow_dissector,
209 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
210 					       target_container);
211 		tp->src = key->tp_src;
212 		tp->dst = key->tp_dst;
213 	}
214 }
215 
216 static enum flow_dissect_ret
217 __skb_flow_dissect_mpls(const struct sk_buff *skb,
218 			struct flow_dissector *flow_dissector,
219 			void *target_container, void *data, int nhoff, int hlen)
220 {
221 	struct flow_dissector_key_keyid *key_keyid;
222 	struct mpls_label *hdr, _hdr[2];
223 	u32 entry, label;
224 
225 	if (!dissector_uses_key(flow_dissector,
226 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
227 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
228 		return FLOW_DISSECT_RET_OUT_GOOD;
229 
230 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
231 				   hlen, &_hdr);
232 	if (!hdr)
233 		return FLOW_DISSECT_RET_OUT_BAD;
234 
235 	entry = ntohl(hdr[0].entry);
236 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
237 
238 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
239 		struct flow_dissector_key_mpls *key_mpls;
240 
241 		key_mpls = skb_flow_dissector_target(flow_dissector,
242 						     FLOW_DISSECTOR_KEY_MPLS,
243 						     target_container);
244 		key_mpls->mpls_label = label;
245 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
246 					>> MPLS_LS_TTL_SHIFT;
247 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
248 					>> MPLS_LS_TC_SHIFT;
249 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
250 					>> MPLS_LS_S_SHIFT;
251 	}
252 
253 	if (label == MPLS_LABEL_ENTROPY) {
254 		key_keyid = skb_flow_dissector_target(flow_dissector,
255 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
256 						      target_container);
257 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
258 	}
259 	return FLOW_DISSECT_RET_OUT_GOOD;
260 }
261 
262 static enum flow_dissect_ret
263 __skb_flow_dissect_arp(const struct sk_buff *skb,
264 		       struct flow_dissector *flow_dissector,
265 		       void *target_container, void *data, int nhoff, int hlen)
266 {
267 	struct flow_dissector_key_arp *key_arp;
268 	struct {
269 		unsigned char ar_sha[ETH_ALEN];
270 		unsigned char ar_sip[4];
271 		unsigned char ar_tha[ETH_ALEN];
272 		unsigned char ar_tip[4];
273 	} *arp_eth, _arp_eth;
274 	const struct arphdr *arp;
275 	struct arphdr _arp;
276 
277 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
278 		return FLOW_DISSECT_RET_OUT_GOOD;
279 
280 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
281 				   hlen, &_arp);
282 	if (!arp)
283 		return FLOW_DISSECT_RET_OUT_BAD;
284 
285 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
286 	    arp->ar_pro != htons(ETH_P_IP) ||
287 	    arp->ar_hln != ETH_ALEN ||
288 	    arp->ar_pln != 4 ||
289 	    (arp->ar_op != htons(ARPOP_REPLY) &&
290 	     arp->ar_op != htons(ARPOP_REQUEST)))
291 		return FLOW_DISSECT_RET_OUT_BAD;
292 
293 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
294 				       sizeof(_arp_eth), data,
295 				       hlen, &_arp_eth);
296 	if (!arp_eth)
297 		return FLOW_DISSECT_RET_OUT_BAD;
298 
299 	key_arp = skb_flow_dissector_target(flow_dissector,
300 					    FLOW_DISSECTOR_KEY_ARP,
301 					    target_container);
302 
303 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
304 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
305 
306 	/* Only store the lower byte of the opcode;
307 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
308 	 */
309 	key_arp->op = ntohs(arp->ar_op) & 0xff;
310 
311 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
312 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
313 
314 	return FLOW_DISSECT_RET_OUT_GOOD;
315 }
316 
317 static enum flow_dissect_ret
318 __skb_flow_dissect_gre(const struct sk_buff *skb,
319 		       struct flow_dissector_key_control *key_control,
320 		       struct flow_dissector *flow_dissector,
321 		       void *target_container, void *data,
322 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
323 		       unsigned int flags)
324 {
325 	struct flow_dissector_key_keyid *key_keyid;
326 	struct gre_base_hdr *hdr, _hdr;
327 	int offset = 0;
328 	u16 gre_ver;
329 
330 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
331 				   data, *p_hlen, &_hdr);
332 	if (!hdr)
333 		return FLOW_DISSECT_RET_OUT_BAD;
334 
335 	/* Only look inside GRE without routing */
336 	if (hdr->flags & GRE_ROUTING)
337 		return FLOW_DISSECT_RET_OUT_GOOD;
338 
339 	/* Only look inside GRE for version 0 and 1 */
340 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
341 	if (gre_ver > 1)
342 		return FLOW_DISSECT_RET_OUT_GOOD;
343 
344 	*p_proto = hdr->protocol;
345 	if (gre_ver) {
346 		/* Version1 must be PPTP, and check the flags */
347 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
348 			return FLOW_DISSECT_RET_OUT_GOOD;
349 	}
350 
351 	offset += sizeof(struct gre_base_hdr);
352 
353 	if (hdr->flags & GRE_CSUM)
354 		offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
355 			  sizeof(((struct gre_full_hdr *) 0)->reserved1);
356 
357 	if (hdr->flags & GRE_KEY) {
358 		const __be32 *keyid;
359 		__be32 _keyid;
360 
361 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
362 					     sizeof(_keyid),
363 					     data, *p_hlen, &_keyid);
364 		if (!keyid)
365 			return FLOW_DISSECT_RET_OUT_BAD;
366 
367 		if (dissector_uses_key(flow_dissector,
368 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
369 			key_keyid = skb_flow_dissector_target(flow_dissector,
370 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
371 							      target_container);
372 			if (gre_ver == 0)
373 				key_keyid->keyid = *keyid;
374 			else
375 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
376 		}
377 		offset += sizeof(((struct gre_full_hdr *) 0)->key);
378 	}
379 
380 	if (hdr->flags & GRE_SEQ)
381 		offset += sizeof(((struct pptp_gre_header *) 0)->seq);
382 
383 	if (gre_ver == 0) {
384 		if (*p_proto == htons(ETH_P_TEB)) {
385 			const struct ethhdr *eth;
386 			struct ethhdr _eth;
387 
388 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
389 						   sizeof(_eth),
390 						   data, *p_hlen, &_eth);
391 			if (!eth)
392 				return FLOW_DISSECT_RET_OUT_BAD;
393 			*p_proto = eth->h_proto;
394 			offset += sizeof(*eth);
395 
396 			/* Cap headers that we access via pointers at the
397 			 * end of the Ethernet header as our maximum alignment
398 			 * at that point is only 2 bytes.
399 			 */
400 			if (NET_IP_ALIGN)
401 				*p_hlen = *p_nhoff + offset;
402 		}
403 	} else { /* version 1, must be PPTP */
404 		u8 _ppp_hdr[PPP_HDRLEN];
405 		u8 *ppp_hdr;
406 
407 		if (hdr->flags & GRE_ACK)
408 			offset += sizeof(((struct pptp_gre_header *) 0)->ack);
409 
410 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
411 					       sizeof(_ppp_hdr),
412 					       data, *p_hlen, _ppp_hdr);
413 		if (!ppp_hdr)
414 			return FLOW_DISSECT_RET_OUT_BAD;
415 
416 		switch (PPP_PROTOCOL(ppp_hdr)) {
417 		case PPP_IP:
418 			*p_proto = htons(ETH_P_IP);
419 			break;
420 		case PPP_IPV6:
421 			*p_proto = htons(ETH_P_IPV6);
422 			break;
423 		default:
424 			/* Could probably catch some more like MPLS */
425 			break;
426 		}
427 
428 		offset += PPP_HDRLEN;
429 	}
430 
431 	*p_nhoff += offset;
432 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
433 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
434 		return FLOW_DISSECT_RET_OUT_GOOD;
435 
436 	return FLOW_DISSECT_RET_PROTO_AGAIN;
437 }
438 
439 static void
440 __skb_flow_dissect_tcp(const struct sk_buff *skb,
441 		       struct flow_dissector *flow_dissector,
442 		       void *target_container, void *data, int thoff, int hlen)
443 {
444 	struct flow_dissector_key_tcp *key_tcp;
445 	struct tcphdr *th, _th;
446 
447 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
448 		return;
449 
450 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
451 	if (!th)
452 		return;
453 
454 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
455 		return;
456 
457 	key_tcp = skb_flow_dissector_target(flow_dissector,
458 					    FLOW_DISSECTOR_KEY_TCP,
459 					    target_container);
460 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
461 }
462 
463 static void
464 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
465 			struct flow_dissector *flow_dissector,
466 			void *target_container, void *data, const struct iphdr *iph)
467 {
468 	struct flow_dissector_key_ip *key_ip;
469 
470 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
471 		return;
472 
473 	key_ip = skb_flow_dissector_target(flow_dissector,
474 					   FLOW_DISSECTOR_KEY_IP,
475 					   target_container);
476 	key_ip->tos = iph->tos;
477 	key_ip->ttl = iph->ttl;
478 }
479 
480 static void
481 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
482 			struct flow_dissector *flow_dissector,
483 			void *target_container, void *data, const struct ipv6hdr *iph)
484 {
485 	struct flow_dissector_key_ip *key_ip;
486 
487 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
488 		return;
489 
490 	key_ip = skb_flow_dissector_target(flow_dissector,
491 					   FLOW_DISSECTOR_KEY_IP,
492 					   target_container);
493 	key_ip->tos = ipv6_get_dsfield(iph);
494 	key_ip->ttl = iph->hop_limit;
495 }
496 
497 /* Maximum number of protocol headers that can be parsed in
498  * __skb_flow_dissect
499  */
500 #define MAX_FLOW_DISSECT_HDRS	15
501 
502 static bool skb_flow_dissect_allowed(int *num_hdrs)
503 {
504 	++*num_hdrs;
505 
506 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
507 }
508 
509 /**
510  * __skb_flow_dissect - extract the flow_keys struct and return it
511  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
512  * @flow_dissector: list of keys to dissect
513  * @target_container: target structure to put dissected values into
514  * @data: raw buffer pointer to the packet, if NULL use skb->data
515  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
516  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
517  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
518  *
519  * The function will try to retrieve individual keys into target specified
520  * by flow_dissector from either the skbuff or a raw buffer specified by the
521  * rest parameters.
522  *
523  * Caller must take care of zeroing target container memory.
524  */
525 bool __skb_flow_dissect(const struct sk_buff *skb,
526 			struct flow_dissector *flow_dissector,
527 			void *target_container,
528 			void *data, __be16 proto, int nhoff, int hlen,
529 			unsigned int flags)
530 {
531 	struct flow_dissector_key_control *key_control;
532 	struct flow_dissector_key_basic *key_basic;
533 	struct flow_dissector_key_addrs *key_addrs;
534 	struct flow_dissector_key_ports *key_ports;
535 	struct flow_dissector_key_icmp *key_icmp;
536 	struct flow_dissector_key_tags *key_tags;
537 	struct flow_dissector_key_vlan *key_vlan;
538 	enum flow_dissect_ret fdret;
539 	bool skip_vlan = false;
540 	int num_hdrs = 0;
541 	u8 ip_proto = 0;
542 	bool ret;
543 
544 	if (!data) {
545 		data = skb->data;
546 		proto = skb_vlan_tag_present(skb) ?
547 			 skb->vlan_proto : skb->protocol;
548 		nhoff = skb_network_offset(skb);
549 		hlen = skb_headlen(skb);
550 #if IS_ENABLED(CONFIG_NET_DSA)
551 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
552 			const struct dsa_device_ops *ops;
553 			int offset;
554 
555 			ops = skb->dev->dsa_ptr->tag_ops;
556 			if (ops->flow_dissect &&
557 			    !ops->flow_dissect(skb, &proto, &offset)) {
558 				hlen -= offset;
559 				nhoff += offset;
560 			}
561 		}
562 #endif
563 	}
564 
565 	/* It is ensured by skb_flow_dissector_init() that control key will
566 	 * be always present.
567 	 */
568 	key_control = skb_flow_dissector_target(flow_dissector,
569 						FLOW_DISSECTOR_KEY_CONTROL,
570 						target_container);
571 
572 	/* It is ensured by skb_flow_dissector_init() that basic key will
573 	 * be always present.
574 	 */
575 	key_basic = skb_flow_dissector_target(flow_dissector,
576 					      FLOW_DISSECTOR_KEY_BASIC,
577 					      target_container);
578 
579 	__skb_flow_dissect_tunnel_info(skb, flow_dissector,
580 				       target_container);
581 
582 	if (dissector_uses_key(flow_dissector,
583 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
584 		struct ethhdr *eth = eth_hdr(skb);
585 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
586 
587 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
588 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
589 							  target_container);
590 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
591 	}
592 
593 proto_again:
594 	fdret = FLOW_DISSECT_RET_CONTINUE;
595 
596 	switch (proto) {
597 	case htons(ETH_P_IP): {
598 		const struct iphdr *iph;
599 		struct iphdr _iph;
600 
601 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
602 		if (!iph || iph->ihl < 5) {
603 			fdret = FLOW_DISSECT_RET_OUT_BAD;
604 			break;
605 		}
606 
607 		nhoff += iph->ihl * 4;
608 
609 		ip_proto = iph->protocol;
610 
611 		if (dissector_uses_key(flow_dissector,
612 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
613 			key_addrs = skb_flow_dissector_target(flow_dissector,
614 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
615 							      target_container);
616 
617 			memcpy(&key_addrs->v4addrs, &iph->saddr,
618 			       sizeof(key_addrs->v4addrs));
619 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
620 		}
621 
622 		if (ip_is_fragment(iph)) {
623 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
624 
625 			if (iph->frag_off & htons(IP_OFFSET)) {
626 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
627 				break;
628 			} else {
629 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
630 				if (!(flags &
631 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
632 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
633 					break;
634 				}
635 			}
636 		}
637 
638 		__skb_flow_dissect_ipv4(skb, flow_dissector,
639 					target_container, data, iph);
640 
641 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
642 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
643 			break;
644 		}
645 
646 		break;
647 	}
648 	case htons(ETH_P_IPV6): {
649 		const struct ipv6hdr *iph;
650 		struct ipv6hdr _iph;
651 
652 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
653 		if (!iph) {
654 			fdret = FLOW_DISSECT_RET_OUT_BAD;
655 			break;
656 		}
657 
658 		ip_proto = iph->nexthdr;
659 		nhoff += sizeof(struct ipv6hdr);
660 
661 		if (dissector_uses_key(flow_dissector,
662 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
663 			key_addrs = skb_flow_dissector_target(flow_dissector,
664 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
665 							      target_container);
666 
667 			memcpy(&key_addrs->v6addrs, &iph->saddr,
668 			       sizeof(key_addrs->v6addrs));
669 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
670 		}
671 
672 		if ((dissector_uses_key(flow_dissector,
673 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
674 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
675 		    ip6_flowlabel(iph)) {
676 			__be32 flow_label = ip6_flowlabel(iph);
677 
678 			if (dissector_uses_key(flow_dissector,
679 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
680 				key_tags = skb_flow_dissector_target(flow_dissector,
681 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
682 								     target_container);
683 				key_tags->flow_label = ntohl(flow_label);
684 			}
685 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
686 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
687 				break;
688 			}
689 		}
690 
691 		__skb_flow_dissect_ipv6(skb, flow_dissector,
692 					target_container, data, iph);
693 
694 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
695 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
696 
697 		break;
698 	}
699 	case htons(ETH_P_8021AD):
700 	case htons(ETH_P_8021Q): {
701 		const struct vlan_hdr *vlan;
702 		struct vlan_hdr _vlan;
703 		bool vlan_tag_present = skb && skb_vlan_tag_present(skb);
704 
705 		if (vlan_tag_present)
706 			proto = skb->protocol;
707 
708 		if (!vlan_tag_present || eth_type_vlan(skb->protocol)) {
709 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
710 						    data, hlen, &_vlan);
711 			if (!vlan) {
712 				fdret = FLOW_DISSECT_RET_OUT_BAD;
713 				break;
714 			}
715 
716 			proto = vlan->h_vlan_encapsulated_proto;
717 			nhoff += sizeof(*vlan);
718 			if (skip_vlan) {
719 				fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
720 				break;
721 			}
722 		}
723 
724 		skip_vlan = true;
725 		if (dissector_uses_key(flow_dissector,
726 				       FLOW_DISSECTOR_KEY_VLAN)) {
727 			key_vlan = skb_flow_dissector_target(flow_dissector,
728 							     FLOW_DISSECTOR_KEY_VLAN,
729 							     target_container);
730 
731 			if (vlan_tag_present) {
732 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
733 				key_vlan->vlan_priority =
734 					(skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
735 			} else {
736 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
737 					VLAN_VID_MASK;
738 				key_vlan->vlan_priority =
739 					(ntohs(vlan->h_vlan_TCI) &
740 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
741 			}
742 		}
743 
744 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
745 		break;
746 	}
747 	case htons(ETH_P_PPP_SES): {
748 		struct {
749 			struct pppoe_hdr hdr;
750 			__be16 proto;
751 		} *hdr, _hdr;
752 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
753 		if (!hdr) {
754 			fdret = FLOW_DISSECT_RET_OUT_BAD;
755 			break;
756 		}
757 
758 		proto = hdr->proto;
759 		nhoff += PPPOE_SES_HLEN;
760 		switch (proto) {
761 		case htons(PPP_IP):
762 			proto = htons(ETH_P_IP);
763 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
764 			break;
765 		case htons(PPP_IPV6):
766 			proto = htons(ETH_P_IPV6);
767 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
768 			break;
769 		default:
770 			fdret = FLOW_DISSECT_RET_OUT_BAD;
771 			break;
772 		}
773 		break;
774 	}
775 	case htons(ETH_P_TIPC): {
776 		struct tipc_basic_hdr *hdr, _hdr;
777 
778 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
779 					   data, hlen, &_hdr);
780 		if (!hdr) {
781 			fdret = FLOW_DISSECT_RET_OUT_BAD;
782 			break;
783 		}
784 
785 		if (dissector_uses_key(flow_dissector,
786 				       FLOW_DISSECTOR_KEY_TIPC)) {
787 			key_addrs = skb_flow_dissector_target(flow_dissector,
788 							      FLOW_DISSECTOR_KEY_TIPC,
789 							      target_container);
790 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
791 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
792 		}
793 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
794 		break;
795 	}
796 
797 	case htons(ETH_P_MPLS_UC):
798 	case htons(ETH_P_MPLS_MC):
799 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
800 						target_container, data,
801 						nhoff, hlen);
802 		break;
803 	case htons(ETH_P_FCOE):
804 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
805 			fdret = FLOW_DISSECT_RET_OUT_BAD;
806 			break;
807 		}
808 
809 		nhoff += FCOE_HEADER_LEN;
810 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
811 		break;
812 
813 	case htons(ETH_P_ARP):
814 	case htons(ETH_P_RARP):
815 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
816 					       target_container, data,
817 					       nhoff, hlen);
818 		break;
819 
820 	default:
821 		fdret = FLOW_DISSECT_RET_OUT_BAD;
822 		break;
823 	}
824 
825 	/* Process result of proto processing */
826 	switch (fdret) {
827 	case FLOW_DISSECT_RET_OUT_GOOD:
828 		goto out_good;
829 	case FLOW_DISSECT_RET_PROTO_AGAIN:
830 		if (skb_flow_dissect_allowed(&num_hdrs))
831 			goto proto_again;
832 		goto out_good;
833 	case FLOW_DISSECT_RET_CONTINUE:
834 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
835 		break;
836 	case FLOW_DISSECT_RET_OUT_BAD:
837 	default:
838 		goto out_bad;
839 	}
840 
841 ip_proto_again:
842 	fdret = FLOW_DISSECT_RET_CONTINUE;
843 
844 	switch (ip_proto) {
845 	case IPPROTO_GRE:
846 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
847 					       target_container, data,
848 					       &proto, &nhoff, &hlen, flags);
849 		break;
850 
851 	case NEXTHDR_HOP:
852 	case NEXTHDR_ROUTING:
853 	case NEXTHDR_DEST: {
854 		u8 _opthdr[2], *opthdr;
855 
856 		if (proto != htons(ETH_P_IPV6))
857 			break;
858 
859 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
860 					      data, hlen, &_opthdr);
861 		if (!opthdr) {
862 			fdret = FLOW_DISSECT_RET_OUT_BAD;
863 			break;
864 		}
865 
866 		ip_proto = opthdr[0];
867 		nhoff += (opthdr[1] + 1) << 3;
868 
869 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
870 		break;
871 	}
872 	case NEXTHDR_FRAGMENT: {
873 		struct frag_hdr _fh, *fh;
874 
875 		if (proto != htons(ETH_P_IPV6))
876 			break;
877 
878 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
879 					  data, hlen, &_fh);
880 
881 		if (!fh) {
882 			fdret = FLOW_DISSECT_RET_OUT_BAD;
883 			break;
884 		}
885 
886 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
887 
888 		nhoff += sizeof(_fh);
889 		ip_proto = fh->nexthdr;
890 
891 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
892 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
893 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
894 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
895 				break;
896 			}
897 		}
898 
899 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
900 		break;
901 	}
902 	case IPPROTO_IPIP:
903 		proto = htons(ETH_P_IP);
904 
905 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
906 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
907 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
908 			break;
909 		}
910 
911 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
912 		break;
913 
914 	case IPPROTO_IPV6:
915 		proto = htons(ETH_P_IPV6);
916 
917 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
918 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
919 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
920 			break;
921 		}
922 
923 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
924 		break;
925 
926 
927 	case IPPROTO_MPLS:
928 		proto = htons(ETH_P_MPLS_UC);
929 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
930 		break;
931 
932 	case IPPROTO_TCP:
933 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
934 				       data, nhoff, hlen);
935 		break;
936 
937 	default:
938 		break;
939 	}
940 
941 	if (dissector_uses_key(flow_dissector,
942 			       FLOW_DISSECTOR_KEY_PORTS)) {
943 		key_ports = skb_flow_dissector_target(flow_dissector,
944 						      FLOW_DISSECTOR_KEY_PORTS,
945 						      target_container);
946 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
947 							data, hlen);
948 	}
949 
950 	if (dissector_uses_key(flow_dissector,
951 			       FLOW_DISSECTOR_KEY_ICMP)) {
952 		key_icmp = skb_flow_dissector_target(flow_dissector,
953 						     FLOW_DISSECTOR_KEY_ICMP,
954 						     target_container);
955 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
956 	}
957 
958 	/* Process result of IP proto processing */
959 	switch (fdret) {
960 	case FLOW_DISSECT_RET_PROTO_AGAIN:
961 		if (skb_flow_dissect_allowed(&num_hdrs))
962 			goto proto_again;
963 		break;
964 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
965 		if (skb_flow_dissect_allowed(&num_hdrs))
966 			goto ip_proto_again;
967 		break;
968 	case FLOW_DISSECT_RET_OUT_GOOD:
969 	case FLOW_DISSECT_RET_CONTINUE:
970 		break;
971 	case FLOW_DISSECT_RET_OUT_BAD:
972 	default:
973 		goto out_bad;
974 	}
975 
976 out_good:
977 	ret = true;
978 
979 	key_control->thoff = (u16)nhoff;
980 out:
981 	key_basic->n_proto = proto;
982 	key_basic->ip_proto = ip_proto;
983 
984 	return ret;
985 
986 out_bad:
987 	ret = false;
988 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
989 	goto out;
990 }
991 EXPORT_SYMBOL(__skb_flow_dissect);
992 
993 static u32 hashrnd __read_mostly;
994 static __always_inline void __flow_hash_secret_init(void)
995 {
996 	net_get_random_once(&hashrnd, sizeof(hashrnd));
997 }
998 
999 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1000 					     u32 keyval)
1001 {
1002 	return jhash2(words, length, keyval);
1003 }
1004 
1005 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1006 {
1007 	const void *p = flow;
1008 
1009 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1010 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1011 }
1012 
1013 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1014 {
1015 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1016 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1017 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1018 		     sizeof(*flow) - sizeof(flow->addrs));
1019 
1020 	switch (flow->control.addr_type) {
1021 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1022 		diff -= sizeof(flow->addrs.v4addrs);
1023 		break;
1024 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1025 		diff -= sizeof(flow->addrs.v6addrs);
1026 		break;
1027 	case FLOW_DISSECTOR_KEY_TIPC:
1028 		diff -= sizeof(flow->addrs.tipckey);
1029 		break;
1030 	}
1031 	return (sizeof(*flow) - diff) / sizeof(u32);
1032 }
1033 
1034 __be32 flow_get_u32_src(const struct flow_keys *flow)
1035 {
1036 	switch (flow->control.addr_type) {
1037 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1038 		return flow->addrs.v4addrs.src;
1039 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1040 		return (__force __be32)ipv6_addr_hash(
1041 			&flow->addrs.v6addrs.src);
1042 	case FLOW_DISSECTOR_KEY_TIPC:
1043 		return flow->addrs.tipckey.key;
1044 	default:
1045 		return 0;
1046 	}
1047 }
1048 EXPORT_SYMBOL(flow_get_u32_src);
1049 
1050 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1051 {
1052 	switch (flow->control.addr_type) {
1053 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1054 		return flow->addrs.v4addrs.dst;
1055 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1056 		return (__force __be32)ipv6_addr_hash(
1057 			&flow->addrs.v6addrs.dst);
1058 	default:
1059 		return 0;
1060 	}
1061 }
1062 EXPORT_SYMBOL(flow_get_u32_dst);
1063 
1064 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1065 {
1066 	int addr_diff, i;
1067 
1068 	switch (keys->control.addr_type) {
1069 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1070 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1071 			    (__force u32)keys->addrs.v4addrs.src;
1072 		if ((addr_diff < 0) ||
1073 		    (addr_diff == 0 &&
1074 		     ((__force u16)keys->ports.dst <
1075 		      (__force u16)keys->ports.src))) {
1076 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1077 			swap(keys->ports.src, keys->ports.dst);
1078 		}
1079 		break;
1080 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1081 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1082 				   &keys->addrs.v6addrs.src,
1083 				   sizeof(keys->addrs.v6addrs.dst));
1084 		if ((addr_diff < 0) ||
1085 		    (addr_diff == 0 &&
1086 		     ((__force u16)keys->ports.dst <
1087 		      (__force u16)keys->ports.src))) {
1088 			for (i = 0; i < 4; i++)
1089 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1090 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1091 			swap(keys->ports.src, keys->ports.dst);
1092 		}
1093 		break;
1094 	}
1095 }
1096 
1097 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1098 {
1099 	u32 hash;
1100 
1101 	__flow_hash_consistentify(keys);
1102 
1103 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1104 				 flow_keys_hash_length(keys), keyval);
1105 	if (!hash)
1106 		hash = 1;
1107 
1108 	return hash;
1109 }
1110 
1111 u32 flow_hash_from_keys(struct flow_keys *keys)
1112 {
1113 	__flow_hash_secret_init();
1114 	return __flow_hash_from_keys(keys, hashrnd);
1115 }
1116 EXPORT_SYMBOL(flow_hash_from_keys);
1117 
1118 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1119 				  struct flow_keys *keys, u32 keyval)
1120 {
1121 	skb_flow_dissect_flow_keys(skb, keys,
1122 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1123 
1124 	return __flow_hash_from_keys(keys, keyval);
1125 }
1126 
1127 struct _flow_keys_digest_data {
1128 	__be16	n_proto;
1129 	u8	ip_proto;
1130 	u8	padding;
1131 	__be32	ports;
1132 	__be32	src;
1133 	__be32	dst;
1134 };
1135 
1136 void make_flow_keys_digest(struct flow_keys_digest *digest,
1137 			   const struct flow_keys *flow)
1138 {
1139 	struct _flow_keys_digest_data *data =
1140 	    (struct _flow_keys_digest_data *)digest;
1141 
1142 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1143 
1144 	memset(digest, 0, sizeof(*digest));
1145 
1146 	data->n_proto = flow->basic.n_proto;
1147 	data->ip_proto = flow->basic.ip_proto;
1148 	data->ports = flow->ports.ports;
1149 	data->src = flow->addrs.v4addrs.src;
1150 	data->dst = flow->addrs.v4addrs.dst;
1151 }
1152 EXPORT_SYMBOL(make_flow_keys_digest);
1153 
1154 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1155 
1156 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1157 {
1158 	struct flow_keys keys;
1159 
1160 	__flow_hash_secret_init();
1161 
1162 	memset(&keys, 0, sizeof(keys));
1163 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1164 			   NULL, 0, 0, 0,
1165 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1166 
1167 	return __flow_hash_from_keys(&keys, hashrnd);
1168 }
1169 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1170 
1171 /**
1172  * __skb_get_hash: calculate a flow hash
1173  * @skb: sk_buff to calculate flow hash from
1174  *
1175  * This function calculates a flow hash based on src/dst addresses
1176  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1177  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1178  * if hash is a canonical 4-tuple hash over transport ports.
1179  */
1180 void __skb_get_hash(struct sk_buff *skb)
1181 {
1182 	struct flow_keys keys;
1183 	u32 hash;
1184 
1185 	__flow_hash_secret_init();
1186 
1187 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1188 
1189 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1190 }
1191 EXPORT_SYMBOL(__skb_get_hash);
1192 
1193 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1194 {
1195 	struct flow_keys keys;
1196 
1197 	return ___skb_get_hash(skb, &keys, perturb);
1198 }
1199 EXPORT_SYMBOL(skb_get_hash_perturb);
1200 
1201 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1202 		   const struct flow_keys *keys, int hlen)
1203 {
1204 	u32 poff = keys->control.thoff;
1205 
1206 	/* skip L4 headers for fragments after the first */
1207 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1208 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1209 		return poff;
1210 
1211 	switch (keys->basic.ip_proto) {
1212 	case IPPROTO_TCP: {
1213 		/* access doff as u8 to avoid unaligned access */
1214 		const u8 *doff;
1215 		u8 _doff;
1216 
1217 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1218 					    data, hlen, &_doff);
1219 		if (!doff)
1220 			return poff;
1221 
1222 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1223 		break;
1224 	}
1225 	case IPPROTO_UDP:
1226 	case IPPROTO_UDPLITE:
1227 		poff += sizeof(struct udphdr);
1228 		break;
1229 	/* For the rest, we do not really care about header
1230 	 * extensions at this point for now.
1231 	 */
1232 	case IPPROTO_ICMP:
1233 		poff += sizeof(struct icmphdr);
1234 		break;
1235 	case IPPROTO_ICMPV6:
1236 		poff += sizeof(struct icmp6hdr);
1237 		break;
1238 	case IPPROTO_IGMP:
1239 		poff += sizeof(struct igmphdr);
1240 		break;
1241 	case IPPROTO_DCCP:
1242 		poff += sizeof(struct dccp_hdr);
1243 		break;
1244 	case IPPROTO_SCTP:
1245 		poff += sizeof(struct sctphdr);
1246 		break;
1247 	}
1248 
1249 	return poff;
1250 }
1251 
1252 /**
1253  * skb_get_poff - get the offset to the payload
1254  * @skb: sk_buff to get the payload offset from
1255  *
1256  * The function will get the offset to the payload as far as it could
1257  * be dissected.  The main user is currently BPF, so that we can dynamically
1258  * truncate packets without needing to push actual payload to the user
1259  * space and can analyze headers only, instead.
1260  */
1261 u32 skb_get_poff(const struct sk_buff *skb)
1262 {
1263 	struct flow_keys keys;
1264 
1265 	if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
1266 		return 0;
1267 
1268 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1269 }
1270 
1271 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1272 {
1273 	memset(keys, 0, sizeof(*keys));
1274 
1275 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1276 	    sizeof(keys->addrs.v6addrs.src));
1277 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1278 	    sizeof(keys->addrs.v6addrs.dst));
1279 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1280 	keys->ports.src = fl6->fl6_sport;
1281 	keys->ports.dst = fl6->fl6_dport;
1282 	keys->keyid.keyid = fl6->fl6_gre_key;
1283 	keys->tags.flow_label = (__force u32)fl6->flowlabel;
1284 	keys->basic.ip_proto = fl6->flowi6_proto;
1285 
1286 	return flow_hash_from_keys(keys);
1287 }
1288 EXPORT_SYMBOL(__get_hash_from_flowi6);
1289 
1290 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys)
1291 {
1292 	memset(keys, 0, sizeof(*keys));
1293 
1294 	keys->addrs.v4addrs.src = fl4->saddr;
1295 	keys->addrs.v4addrs.dst = fl4->daddr;
1296 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1297 	keys->ports.src = fl4->fl4_sport;
1298 	keys->ports.dst = fl4->fl4_dport;
1299 	keys->keyid.keyid = fl4->fl4_gre_key;
1300 	keys->basic.ip_proto = fl4->flowi4_proto;
1301 
1302 	return flow_hash_from_keys(keys);
1303 }
1304 EXPORT_SYMBOL(__get_hash_from_flowi4);
1305 
1306 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1307 	{
1308 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1309 		.offset = offsetof(struct flow_keys, control),
1310 	},
1311 	{
1312 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1313 		.offset = offsetof(struct flow_keys, basic),
1314 	},
1315 	{
1316 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1317 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1318 	},
1319 	{
1320 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1321 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1322 	},
1323 	{
1324 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1325 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1326 	},
1327 	{
1328 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1329 		.offset = offsetof(struct flow_keys, ports),
1330 	},
1331 	{
1332 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1333 		.offset = offsetof(struct flow_keys, vlan),
1334 	},
1335 	{
1336 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1337 		.offset = offsetof(struct flow_keys, tags),
1338 	},
1339 	{
1340 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1341 		.offset = offsetof(struct flow_keys, keyid),
1342 	},
1343 };
1344 
1345 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1346 	{
1347 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1348 		.offset = offsetof(struct flow_keys, control),
1349 	},
1350 	{
1351 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1352 		.offset = offsetof(struct flow_keys, basic),
1353 	},
1354 	{
1355 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1356 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1357 	},
1358 	{
1359 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1360 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1361 	},
1362 	{
1363 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1364 		.offset = offsetof(struct flow_keys, ports),
1365 	},
1366 };
1367 
1368 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
1369 	{
1370 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1371 		.offset = offsetof(struct flow_keys, control),
1372 	},
1373 	{
1374 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1375 		.offset = offsetof(struct flow_keys, basic),
1376 	},
1377 };
1378 
1379 struct flow_dissector flow_keys_dissector __read_mostly;
1380 EXPORT_SYMBOL(flow_keys_dissector);
1381 
1382 struct flow_dissector flow_keys_buf_dissector __read_mostly;
1383 
1384 static int __init init_default_flow_dissectors(void)
1385 {
1386 	skb_flow_dissector_init(&flow_keys_dissector,
1387 				flow_keys_dissector_keys,
1388 				ARRAY_SIZE(flow_keys_dissector_keys));
1389 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1390 				flow_keys_dissector_symmetric_keys,
1391 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1392 	skb_flow_dissector_init(&flow_keys_buf_dissector,
1393 				flow_keys_buf_dissector_keys,
1394 				ARRAY_SIZE(flow_keys_buf_dissector_keys));
1395 	return 0;
1396 }
1397 
1398 core_initcall(init_default_flow_dissectors);
1399