xref: /linux/net/core/flow_dissector.c (revision e9f0878c4b2004ac19581274c1ae4c61ae3ca70e)
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 #include <uapi/linux/batadv_packet.h>
28 
29 static void dissector_set_key(struct flow_dissector *flow_dissector,
30 			      enum flow_dissector_key_id key_id)
31 {
32 	flow_dissector->used_keys |= (1 << key_id);
33 }
34 
35 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
36 			     const struct flow_dissector_key *key,
37 			     unsigned int key_count)
38 {
39 	unsigned int i;
40 
41 	memset(flow_dissector, 0, sizeof(*flow_dissector));
42 
43 	for (i = 0; i < key_count; i++, key++) {
44 		/* User should make sure that every key target offset is withing
45 		 * boundaries of unsigned short.
46 		 */
47 		BUG_ON(key->offset > USHRT_MAX);
48 		BUG_ON(dissector_uses_key(flow_dissector,
49 					  key->key_id));
50 
51 		dissector_set_key(flow_dissector, key->key_id);
52 		flow_dissector->offset[key->key_id] = key->offset;
53 	}
54 
55 	/* Ensure that the dissector always includes control and basic key.
56 	 * That way we are able to avoid handling lack of these in fast path.
57 	 */
58 	BUG_ON(!dissector_uses_key(flow_dissector,
59 				   FLOW_DISSECTOR_KEY_CONTROL));
60 	BUG_ON(!dissector_uses_key(flow_dissector,
61 				   FLOW_DISSECTOR_KEY_BASIC));
62 }
63 EXPORT_SYMBOL(skb_flow_dissector_init);
64 
65 /**
66  * skb_flow_get_be16 - extract be16 entity
67  * @skb: sk_buff to extract from
68  * @poff: offset to extract at
69  * @data: raw buffer pointer to the packet
70  * @hlen: packet header length
71  *
72  * The function will try to retrieve a be32 entity at
73  * offset poff
74  */
75 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
76 				void *data, int hlen)
77 {
78 	__be16 *u, _u;
79 
80 	u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
81 	if (u)
82 		return *u;
83 
84 	return 0;
85 }
86 
87 /**
88  * __skb_flow_get_ports - extract the upper layer ports and return them
89  * @skb: sk_buff to extract the ports from
90  * @thoff: transport header offset
91  * @ip_proto: protocol for which to get port offset
92  * @data: raw buffer pointer to the packet, if NULL use skb->data
93  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
94  *
95  * The function will try to retrieve the ports at offset thoff + poff where poff
96  * is the protocol port offset returned from proto_ports_offset
97  */
98 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
99 			    void *data, int hlen)
100 {
101 	int poff = proto_ports_offset(ip_proto);
102 
103 	if (!data) {
104 		data = skb->data;
105 		hlen = skb_headlen(skb);
106 	}
107 
108 	if (poff >= 0) {
109 		__be32 *ports, _ports;
110 
111 		ports = __skb_header_pointer(skb, thoff + poff,
112 					     sizeof(_ports), data, hlen, &_ports);
113 		if (ports)
114 			return *ports;
115 	}
116 
117 	return 0;
118 }
119 EXPORT_SYMBOL(__skb_flow_get_ports);
120 
121 static void
122 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
123 				   struct flow_dissector *flow_dissector,
124 				   void *target_container)
125 {
126 	struct flow_dissector_key_control *ctrl;
127 
128 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
129 		return;
130 
131 	ctrl = skb_flow_dissector_target(flow_dissector,
132 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
133 					 target_container);
134 	ctrl->addr_type = type;
135 }
136 
137 void
138 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
139 			     struct flow_dissector *flow_dissector,
140 			     void *target_container)
141 {
142 	struct ip_tunnel_info *info;
143 	struct ip_tunnel_key *key;
144 
145 	/* A quick check to see if there might be something to do. */
146 	if (!dissector_uses_key(flow_dissector,
147 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
148 	    !dissector_uses_key(flow_dissector,
149 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
150 	    !dissector_uses_key(flow_dissector,
151 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
152 	    !dissector_uses_key(flow_dissector,
153 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
154 	    !dissector_uses_key(flow_dissector,
155 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
156 	    !dissector_uses_key(flow_dissector,
157 				FLOW_DISSECTOR_KEY_ENC_IP) &&
158 	    !dissector_uses_key(flow_dissector,
159 				FLOW_DISSECTOR_KEY_ENC_OPTS))
160 		return;
161 
162 	info = skb_tunnel_info(skb);
163 	if (!info)
164 		return;
165 
166 	key = &info->key;
167 
168 	switch (ip_tunnel_info_af(info)) {
169 	case AF_INET:
170 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
171 						   flow_dissector,
172 						   target_container);
173 		if (dissector_uses_key(flow_dissector,
174 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
175 			struct flow_dissector_key_ipv4_addrs *ipv4;
176 
177 			ipv4 = skb_flow_dissector_target(flow_dissector,
178 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
179 							 target_container);
180 			ipv4->src = key->u.ipv4.src;
181 			ipv4->dst = key->u.ipv4.dst;
182 		}
183 		break;
184 	case AF_INET6:
185 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
186 						   flow_dissector,
187 						   target_container);
188 		if (dissector_uses_key(flow_dissector,
189 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
190 			struct flow_dissector_key_ipv6_addrs *ipv6;
191 
192 			ipv6 = skb_flow_dissector_target(flow_dissector,
193 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
194 							 target_container);
195 			ipv6->src = key->u.ipv6.src;
196 			ipv6->dst = key->u.ipv6.dst;
197 		}
198 		break;
199 	}
200 
201 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
202 		struct flow_dissector_key_keyid *keyid;
203 
204 		keyid = skb_flow_dissector_target(flow_dissector,
205 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
206 						  target_container);
207 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
208 	}
209 
210 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
211 		struct flow_dissector_key_ports *tp;
212 
213 		tp = skb_flow_dissector_target(flow_dissector,
214 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
215 					       target_container);
216 		tp->src = key->tp_src;
217 		tp->dst = key->tp_dst;
218 	}
219 
220 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
221 		struct flow_dissector_key_ip *ip;
222 
223 		ip = skb_flow_dissector_target(flow_dissector,
224 					       FLOW_DISSECTOR_KEY_ENC_IP,
225 					       target_container);
226 		ip->tos = key->tos;
227 		ip->ttl = key->ttl;
228 	}
229 
230 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
231 		struct flow_dissector_key_enc_opts *enc_opt;
232 
233 		enc_opt = skb_flow_dissector_target(flow_dissector,
234 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
235 						    target_container);
236 
237 		if (info->options_len) {
238 			enc_opt->len = info->options_len;
239 			ip_tunnel_info_opts_get(enc_opt->data, info);
240 			enc_opt->dst_opt_type = info->key.tun_flags &
241 						TUNNEL_OPTIONS_PRESENT;
242 		}
243 	}
244 }
245 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
246 
247 static enum flow_dissect_ret
248 __skb_flow_dissect_mpls(const struct sk_buff *skb,
249 			struct flow_dissector *flow_dissector,
250 			void *target_container, void *data, int nhoff, int hlen)
251 {
252 	struct flow_dissector_key_keyid *key_keyid;
253 	struct mpls_label *hdr, _hdr[2];
254 	u32 entry, label;
255 
256 	if (!dissector_uses_key(flow_dissector,
257 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
258 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
259 		return FLOW_DISSECT_RET_OUT_GOOD;
260 
261 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
262 				   hlen, &_hdr);
263 	if (!hdr)
264 		return FLOW_DISSECT_RET_OUT_BAD;
265 
266 	entry = ntohl(hdr[0].entry);
267 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
268 
269 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
270 		struct flow_dissector_key_mpls *key_mpls;
271 
272 		key_mpls = skb_flow_dissector_target(flow_dissector,
273 						     FLOW_DISSECTOR_KEY_MPLS,
274 						     target_container);
275 		key_mpls->mpls_label = label;
276 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
277 					>> MPLS_LS_TTL_SHIFT;
278 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
279 					>> MPLS_LS_TC_SHIFT;
280 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
281 					>> MPLS_LS_S_SHIFT;
282 	}
283 
284 	if (label == MPLS_LABEL_ENTROPY) {
285 		key_keyid = skb_flow_dissector_target(flow_dissector,
286 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
287 						      target_container);
288 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
289 	}
290 	return FLOW_DISSECT_RET_OUT_GOOD;
291 }
292 
293 static enum flow_dissect_ret
294 __skb_flow_dissect_arp(const struct sk_buff *skb,
295 		       struct flow_dissector *flow_dissector,
296 		       void *target_container, void *data, int nhoff, int hlen)
297 {
298 	struct flow_dissector_key_arp *key_arp;
299 	struct {
300 		unsigned char ar_sha[ETH_ALEN];
301 		unsigned char ar_sip[4];
302 		unsigned char ar_tha[ETH_ALEN];
303 		unsigned char ar_tip[4];
304 	} *arp_eth, _arp_eth;
305 	const struct arphdr *arp;
306 	struct arphdr _arp;
307 
308 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
309 		return FLOW_DISSECT_RET_OUT_GOOD;
310 
311 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
312 				   hlen, &_arp);
313 	if (!arp)
314 		return FLOW_DISSECT_RET_OUT_BAD;
315 
316 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
317 	    arp->ar_pro != htons(ETH_P_IP) ||
318 	    arp->ar_hln != ETH_ALEN ||
319 	    arp->ar_pln != 4 ||
320 	    (arp->ar_op != htons(ARPOP_REPLY) &&
321 	     arp->ar_op != htons(ARPOP_REQUEST)))
322 		return FLOW_DISSECT_RET_OUT_BAD;
323 
324 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
325 				       sizeof(_arp_eth), data,
326 				       hlen, &_arp_eth);
327 	if (!arp_eth)
328 		return FLOW_DISSECT_RET_OUT_BAD;
329 
330 	key_arp = skb_flow_dissector_target(flow_dissector,
331 					    FLOW_DISSECTOR_KEY_ARP,
332 					    target_container);
333 
334 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
335 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
336 
337 	/* Only store the lower byte of the opcode;
338 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
339 	 */
340 	key_arp->op = ntohs(arp->ar_op) & 0xff;
341 
342 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
343 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
344 
345 	return FLOW_DISSECT_RET_OUT_GOOD;
346 }
347 
348 static enum flow_dissect_ret
349 __skb_flow_dissect_gre(const struct sk_buff *skb,
350 		       struct flow_dissector_key_control *key_control,
351 		       struct flow_dissector *flow_dissector,
352 		       void *target_container, void *data,
353 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
354 		       unsigned int flags)
355 {
356 	struct flow_dissector_key_keyid *key_keyid;
357 	struct gre_base_hdr *hdr, _hdr;
358 	int offset = 0;
359 	u16 gre_ver;
360 
361 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
362 				   data, *p_hlen, &_hdr);
363 	if (!hdr)
364 		return FLOW_DISSECT_RET_OUT_BAD;
365 
366 	/* Only look inside GRE without routing */
367 	if (hdr->flags & GRE_ROUTING)
368 		return FLOW_DISSECT_RET_OUT_GOOD;
369 
370 	/* Only look inside GRE for version 0 and 1 */
371 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
372 	if (gre_ver > 1)
373 		return FLOW_DISSECT_RET_OUT_GOOD;
374 
375 	*p_proto = hdr->protocol;
376 	if (gre_ver) {
377 		/* Version1 must be PPTP, and check the flags */
378 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
379 			return FLOW_DISSECT_RET_OUT_GOOD;
380 	}
381 
382 	offset += sizeof(struct gre_base_hdr);
383 
384 	if (hdr->flags & GRE_CSUM)
385 		offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
386 			  sizeof(((struct gre_full_hdr *) 0)->reserved1);
387 
388 	if (hdr->flags & GRE_KEY) {
389 		const __be32 *keyid;
390 		__be32 _keyid;
391 
392 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
393 					     sizeof(_keyid),
394 					     data, *p_hlen, &_keyid);
395 		if (!keyid)
396 			return FLOW_DISSECT_RET_OUT_BAD;
397 
398 		if (dissector_uses_key(flow_dissector,
399 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
400 			key_keyid = skb_flow_dissector_target(flow_dissector,
401 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
402 							      target_container);
403 			if (gre_ver == 0)
404 				key_keyid->keyid = *keyid;
405 			else
406 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
407 		}
408 		offset += sizeof(((struct gre_full_hdr *) 0)->key);
409 	}
410 
411 	if (hdr->flags & GRE_SEQ)
412 		offset += sizeof(((struct pptp_gre_header *) 0)->seq);
413 
414 	if (gre_ver == 0) {
415 		if (*p_proto == htons(ETH_P_TEB)) {
416 			const struct ethhdr *eth;
417 			struct ethhdr _eth;
418 
419 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
420 						   sizeof(_eth),
421 						   data, *p_hlen, &_eth);
422 			if (!eth)
423 				return FLOW_DISSECT_RET_OUT_BAD;
424 			*p_proto = eth->h_proto;
425 			offset += sizeof(*eth);
426 
427 			/* Cap headers that we access via pointers at the
428 			 * end of the Ethernet header as our maximum alignment
429 			 * at that point is only 2 bytes.
430 			 */
431 			if (NET_IP_ALIGN)
432 				*p_hlen = *p_nhoff + offset;
433 		}
434 	} else { /* version 1, must be PPTP */
435 		u8 _ppp_hdr[PPP_HDRLEN];
436 		u8 *ppp_hdr;
437 
438 		if (hdr->flags & GRE_ACK)
439 			offset += sizeof(((struct pptp_gre_header *) 0)->ack);
440 
441 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
442 					       sizeof(_ppp_hdr),
443 					       data, *p_hlen, _ppp_hdr);
444 		if (!ppp_hdr)
445 			return FLOW_DISSECT_RET_OUT_BAD;
446 
447 		switch (PPP_PROTOCOL(ppp_hdr)) {
448 		case PPP_IP:
449 			*p_proto = htons(ETH_P_IP);
450 			break;
451 		case PPP_IPV6:
452 			*p_proto = htons(ETH_P_IPV6);
453 			break;
454 		default:
455 			/* Could probably catch some more like MPLS */
456 			break;
457 		}
458 
459 		offset += PPP_HDRLEN;
460 	}
461 
462 	*p_nhoff += offset;
463 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
464 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
465 		return FLOW_DISSECT_RET_OUT_GOOD;
466 
467 	return FLOW_DISSECT_RET_PROTO_AGAIN;
468 }
469 
470 /**
471  * __skb_flow_dissect_batadv() - dissect batman-adv header
472  * @skb: sk_buff to with the batman-adv header
473  * @key_control: flow dissectors control key
474  * @data: raw buffer pointer to the packet, if NULL use skb->data
475  * @p_proto: pointer used to update the protocol to process next
476  * @p_nhoff: pointer used to update inner network header offset
477  * @hlen: packet header length
478  * @flags: any combination of FLOW_DISSECTOR_F_*
479  *
480  * ETH_P_BATMAN packets are tried to be dissected. Only
481  * &struct batadv_unicast packets are actually processed because they contain an
482  * inner ethernet header and are usually followed by actual network header. This
483  * allows the flow dissector to continue processing the packet.
484  *
485  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
486  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
487  *  otherwise FLOW_DISSECT_RET_OUT_BAD
488  */
489 static enum flow_dissect_ret
490 __skb_flow_dissect_batadv(const struct sk_buff *skb,
491 			  struct flow_dissector_key_control *key_control,
492 			  void *data, __be16 *p_proto, int *p_nhoff, int hlen,
493 			  unsigned int flags)
494 {
495 	struct {
496 		struct batadv_unicast_packet batadv_unicast;
497 		struct ethhdr eth;
498 	} *hdr, _hdr;
499 
500 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
501 				   &_hdr);
502 	if (!hdr)
503 		return FLOW_DISSECT_RET_OUT_BAD;
504 
505 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
506 		return FLOW_DISSECT_RET_OUT_BAD;
507 
508 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
509 		return FLOW_DISSECT_RET_OUT_BAD;
510 
511 	*p_proto = hdr->eth.h_proto;
512 	*p_nhoff += sizeof(*hdr);
513 
514 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
515 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
516 		return FLOW_DISSECT_RET_OUT_GOOD;
517 
518 	return FLOW_DISSECT_RET_PROTO_AGAIN;
519 }
520 
521 static void
522 __skb_flow_dissect_tcp(const struct sk_buff *skb,
523 		       struct flow_dissector *flow_dissector,
524 		       void *target_container, void *data, int thoff, int hlen)
525 {
526 	struct flow_dissector_key_tcp *key_tcp;
527 	struct tcphdr *th, _th;
528 
529 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
530 		return;
531 
532 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
533 	if (!th)
534 		return;
535 
536 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
537 		return;
538 
539 	key_tcp = skb_flow_dissector_target(flow_dissector,
540 					    FLOW_DISSECTOR_KEY_TCP,
541 					    target_container);
542 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
543 }
544 
545 static void
546 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
547 			struct flow_dissector *flow_dissector,
548 			void *target_container, void *data, const struct iphdr *iph)
549 {
550 	struct flow_dissector_key_ip *key_ip;
551 
552 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
553 		return;
554 
555 	key_ip = skb_flow_dissector_target(flow_dissector,
556 					   FLOW_DISSECTOR_KEY_IP,
557 					   target_container);
558 	key_ip->tos = iph->tos;
559 	key_ip->ttl = iph->ttl;
560 }
561 
562 static void
563 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
564 			struct flow_dissector *flow_dissector,
565 			void *target_container, void *data, const struct ipv6hdr *iph)
566 {
567 	struct flow_dissector_key_ip *key_ip;
568 
569 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
570 		return;
571 
572 	key_ip = skb_flow_dissector_target(flow_dissector,
573 					   FLOW_DISSECTOR_KEY_IP,
574 					   target_container);
575 	key_ip->tos = ipv6_get_dsfield(iph);
576 	key_ip->ttl = iph->hop_limit;
577 }
578 
579 /* Maximum number of protocol headers that can be parsed in
580  * __skb_flow_dissect
581  */
582 #define MAX_FLOW_DISSECT_HDRS	15
583 
584 static bool skb_flow_dissect_allowed(int *num_hdrs)
585 {
586 	++*num_hdrs;
587 
588 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
589 }
590 
591 /**
592  * __skb_flow_dissect - extract the flow_keys struct and return it
593  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
594  * @flow_dissector: list of keys to dissect
595  * @target_container: target structure to put dissected values into
596  * @data: raw buffer pointer to the packet, if NULL use skb->data
597  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
598  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
599  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
600  *
601  * The function will try to retrieve individual keys into target specified
602  * by flow_dissector from either the skbuff or a raw buffer specified by the
603  * rest parameters.
604  *
605  * Caller must take care of zeroing target container memory.
606  */
607 bool __skb_flow_dissect(const struct sk_buff *skb,
608 			struct flow_dissector *flow_dissector,
609 			void *target_container,
610 			void *data, __be16 proto, int nhoff, int hlen,
611 			unsigned int flags)
612 {
613 	struct flow_dissector_key_control *key_control;
614 	struct flow_dissector_key_basic *key_basic;
615 	struct flow_dissector_key_addrs *key_addrs;
616 	struct flow_dissector_key_ports *key_ports;
617 	struct flow_dissector_key_icmp *key_icmp;
618 	struct flow_dissector_key_tags *key_tags;
619 	struct flow_dissector_key_vlan *key_vlan;
620 	enum flow_dissect_ret fdret;
621 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
622 	int num_hdrs = 0;
623 	u8 ip_proto = 0;
624 	bool ret;
625 
626 	if (!data) {
627 		data = skb->data;
628 		proto = skb_vlan_tag_present(skb) ?
629 			 skb->vlan_proto : skb->protocol;
630 		nhoff = skb_network_offset(skb);
631 		hlen = skb_headlen(skb);
632 #if IS_ENABLED(CONFIG_NET_DSA)
633 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
634 			const struct dsa_device_ops *ops;
635 			int offset;
636 
637 			ops = skb->dev->dsa_ptr->tag_ops;
638 			if (ops->flow_dissect &&
639 			    !ops->flow_dissect(skb, &proto, &offset)) {
640 				hlen -= offset;
641 				nhoff += offset;
642 			}
643 		}
644 #endif
645 	}
646 
647 	/* It is ensured by skb_flow_dissector_init() that control key will
648 	 * be always present.
649 	 */
650 	key_control = skb_flow_dissector_target(flow_dissector,
651 						FLOW_DISSECTOR_KEY_CONTROL,
652 						target_container);
653 
654 	/* It is ensured by skb_flow_dissector_init() that basic key will
655 	 * be always present.
656 	 */
657 	key_basic = skb_flow_dissector_target(flow_dissector,
658 					      FLOW_DISSECTOR_KEY_BASIC,
659 					      target_container);
660 
661 	if (dissector_uses_key(flow_dissector,
662 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
663 		struct ethhdr *eth = eth_hdr(skb);
664 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
665 
666 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
667 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
668 							  target_container);
669 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
670 	}
671 
672 proto_again:
673 	fdret = FLOW_DISSECT_RET_CONTINUE;
674 
675 	switch (proto) {
676 	case htons(ETH_P_IP): {
677 		const struct iphdr *iph;
678 		struct iphdr _iph;
679 
680 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
681 		if (!iph || iph->ihl < 5) {
682 			fdret = FLOW_DISSECT_RET_OUT_BAD;
683 			break;
684 		}
685 
686 		nhoff += iph->ihl * 4;
687 
688 		ip_proto = iph->protocol;
689 
690 		if (dissector_uses_key(flow_dissector,
691 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
692 			key_addrs = skb_flow_dissector_target(flow_dissector,
693 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
694 							      target_container);
695 
696 			memcpy(&key_addrs->v4addrs, &iph->saddr,
697 			       sizeof(key_addrs->v4addrs));
698 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
699 		}
700 
701 		if (ip_is_fragment(iph)) {
702 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
703 
704 			if (iph->frag_off & htons(IP_OFFSET)) {
705 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
706 				break;
707 			} else {
708 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
709 				if (!(flags &
710 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
711 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
712 					break;
713 				}
714 			}
715 		}
716 
717 		__skb_flow_dissect_ipv4(skb, flow_dissector,
718 					target_container, data, iph);
719 
720 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
721 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
722 			break;
723 		}
724 
725 		break;
726 	}
727 	case htons(ETH_P_IPV6): {
728 		const struct ipv6hdr *iph;
729 		struct ipv6hdr _iph;
730 
731 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
732 		if (!iph) {
733 			fdret = FLOW_DISSECT_RET_OUT_BAD;
734 			break;
735 		}
736 
737 		ip_proto = iph->nexthdr;
738 		nhoff += sizeof(struct ipv6hdr);
739 
740 		if (dissector_uses_key(flow_dissector,
741 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
742 			key_addrs = skb_flow_dissector_target(flow_dissector,
743 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
744 							      target_container);
745 
746 			memcpy(&key_addrs->v6addrs, &iph->saddr,
747 			       sizeof(key_addrs->v6addrs));
748 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
749 		}
750 
751 		if ((dissector_uses_key(flow_dissector,
752 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
753 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
754 		    ip6_flowlabel(iph)) {
755 			__be32 flow_label = ip6_flowlabel(iph);
756 
757 			if (dissector_uses_key(flow_dissector,
758 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
759 				key_tags = skb_flow_dissector_target(flow_dissector,
760 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
761 								     target_container);
762 				key_tags->flow_label = ntohl(flow_label);
763 			}
764 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
765 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
766 				break;
767 			}
768 		}
769 
770 		__skb_flow_dissect_ipv6(skb, flow_dissector,
771 					target_container, data, iph);
772 
773 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
774 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
775 
776 		break;
777 	}
778 	case htons(ETH_P_8021AD):
779 	case htons(ETH_P_8021Q): {
780 		const struct vlan_hdr *vlan = NULL;
781 		struct vlan_hdr _vlan;
782 		__be16 saved_vlan_tpid = proto;
783 
784 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
785 		    skb && skb_vlan_tag_present(skb)) {
786 			proto = skb->protocol;
787 		} else {
788 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
789 						    data, hlen, &_vlan);
790 			if (!vlan) {
791 				fdret = FLOW_DISSECT_RET_OUT_BAD;
792 				break;
793 			}
794 
795 			proto = vlan->h_vlan_encapsulated_proto;
796 			nhoff += sizeof(*vlan);
797 		}
798 
799 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
800 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
801 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
802 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
803 		} else {
804 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
805 			break;
806 		}
807 
808 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
809 			key_vlan = skb_flow_dissector_target(flow_dissector,
810 							     dissector_vlan,
811 							     target_container);
812 
813 			if (!vlan) {
814 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
815 				key_vlan->vlan_priority =
816 					(skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
817 			} else {
818 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
819 					VLAN_VID_MASK;
820 				key_vlan->vlan_priority =
821 					(ntohs(vlan->h_vlan_TCI) &
822 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
823 			}
824 			key_vlan->vlan_tpid = saved_vlan_tpid;
825 		}
826 
827 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
828 		break;
829 	}
830 	case htons(ETH_P_PPP_SES): {
831 		struct {
832 			struct pppoe_hdr hdr;
833 			__be16 proto;
834 		} *hdr, _hdr;
835 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
836 		if (!hdr) {
837 			fdret = FLOW_DISSECT_RET_OUT_BAD;
838 			break;
839 		}
840 
841 		proto = hdr->proto;
842 		nhoff += PPPOE_SES_HLEN;
843 		switch (proto) {
844 		case htons(PPP_IP):
845 			proto = htons(ETH_P_IP);
846 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
847 			break;
848 		case htons(PPP_IPV6):
849 			proto = htons(ETH_P_IPV6);
850 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
851 			break;
852 		default:
853 			fdret = FLOW_DISSECT_RET_OUT_BAD;
854 			break;
855 		}
856 		break;
857 	}
858 	case htons(ETH_P_TIPC): {
859 		struct tipc_basic_hdr *hdr, _hdr;
860 
861 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
862 					   data, hlen, &_hdr);
863 		if (!hdr) {
864 			fdret = FLOW_DISSECT_RET_OUT_BAD;
865 			break;
866 		}
867 
868 		if (dissector_uses_key(flow_dissector,
869 				       FLOW_DISSECTOR_KEY_TIPC)) {
870 			key_addrs = skb_flow_dissector_target(flow_dissector,
871 							      FLOW_DISSECTOR_KEY_TIPC,
872 							      target_container);
873 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
874 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
875 		}
876 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
877 		break;
878 	}
879 
880 	case htons(ETH_P_MPLS_UC):
881 	case htons(ETH_P_MPLS_MC):
882 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
883 						target_container, data,
884 						nhoff, hlen);
885 		break;
886 	case htons(ETH_P_FCOE):
887 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
888 			fdret = FLOW_DISSECT_RET_OUT_BAD;
889 			break;
890 		}
891 
892 		nhoff += FCOE_HEADER_LEN;
893 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
894 		break;
895 
896 	case htons(ETH_P_ARP):
897 	case htons(ETH_P_RARP):
898 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
899 					       target_container, data,
900 					       nhoff, hlen);
901 		break;
902 
903 	case htons(ETH_P_BATMAN):
904 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
905 						  &proto, &nhoff, hlen, flags);
906 		break;
907 
908 	default:
909 		fdret = FLOW_DISSECT_RET_OUT_BAD;
910 		break;
911 	}
912 
913 	/* Process result of proto processing */
914 	switch (fdret) {
915 	case FLOW_DISSECT_RET_OUT_GOOD:
916 		goto out_good;
917 	case FLOW_DISSECT_RET_PROTO_AGAIN:
918 		if (skb_flow_dissect_allowed(&num_hdrs))
919 			goto proto_again;
920 		goto out_good;
921 	case FLOW_DISSECT_RET_CONTINUE:
922 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
923 		break;
924 	case FLOW_DISSECT_RET_OUT_BAD:
925 	default:
926 		goto out_bad;
927 	}
928 
929 ip_proto_again:
930 	fdret = FLOW_DISSECT_RET_CONTINUE;
931 
932 	switch (ip_proto) {
933 	case IPPROTO_GRE:
934 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
935 					       target_container, data,
936 					       &proto, &nhoff, &hlen, flags);
937 		break;
938 
939 	case NEXTHDR_HOP:
940 	case NEXTHDR_ROUTING:
941 	case NEXTHDR_DEST: {
942 		u8 _opthdr[2], *opthdr;
943 
944 		if (proto != htons(ETH_P_IPV6))
945 			break;
946 
947 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
948 					      data, hlen, &_opthdr);
949 		if (!opthdr) {
950 			fdret = FLOW_DISSECT_RET_OUT_BAD;
951 			break;
952 		}
953 
954 		ip_proto = opthdr[0];
955 		nhoff += (opthdr[1] + 1) << 3;
956 
957 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
958 		break;
959 	}
960 	case NEXTHDR_FRAGMENT: {
961 		struct frag_hdr _fh, *fh;
962 
963 		if (proto != htons(ETH_P_IPV6))
964 			break;
965 
966 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
967 					  data, hlen, &_fh);
968 
969 		if (!fh) {
970 			fdret = FLOW_DISSECT_RET_OUT_BAD;
971 			break;
972 		}
973 
974 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
975 
976 		nhoff += sizeof(_fh);
977 		ip_proto = fh->nexthdr;
978 
979 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
980 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
981 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
982 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
983 				break;
984 			}
985 		}
986 
987 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
988 		break;
989 	}
990 	case IPPROTO_IPIP:
991 		proto = htons(ETH_P_IP);
992 
993 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
994 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
995 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
996 			break;
997 		}
998 
999 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1000 		break;
1001 
1002 	case IPPROTO_IPV6:
1003 		proto = htons(ETH_P_IPV6);
1004 
1005 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1006 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1007 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1008 			break;
1009 		}
1010 
1011 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1012 		break;
1013 
1014 
1015 	case IPPROTO_MPLS:
1016 		proto = htons(ETH_P_MPLS_UC);
1017 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1018 		break;
1019 
1020 	case IPPROTO_TCP:
1021 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1022 				       data, nhoff, hlen);
1023 		break;
1024 
1025 	default:
1026 		break;
1027 	}
1028 
1029 	if (dissector_uses_key(flow_dissector,
1030 			       FLOW_DISSECTOR_KEY_PORTS)) {
1031 		key_ports = skb_flow_dissector_target(flow_dissector,
1032 						      FLOW_DISSECTOR_KEY_PORTS,
1033 						      target_container);
1034 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1035 							data, hlen);
1036 	}
1037 
1038 	if (dissector_uses_key(flow_dissector,
1039 			       FLOW_DISSECTOR_KEY_ICMP)) {
1040 		key_icmp = skb_flow_dissector_target(flow_dissector,
1041 						     FLOW_DISSECTOR_KEY_ICMP,
1042 						     target_container);
1043 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1044 	}
1045 
1046 	/* Process result of IP proto processing */
1047 	switch (fdret) {
1048 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1049 		if (skb_flow_dissect_allowed(&num_hdrs))
1050 			goto proto_again;
1051 		break;
1052 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1053 		if (skb_flow_dissect_allowed(&num_hdrs))
1054 			goto ip_proto_again;
1055 		break;
1056 	case FLOW_DISSECT_RET_OUT_GOOD:
1057 	case FLOW_DISSECT_RET_CONTINUE:
1058 		break;
1059 	case FLOW_DISSECT_RET_OUT_BAD:
1060 	default:
1061 		goto out_bad;
1062 	}
1063 
1064 out_good:
1065 	ret = true;
1066 
1067 out:
1068 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1069 	key_basic->n_proto = proto;
1070 	key_basic->ip_proto = ip_proto;
1071 
1072 	return ret;
1073 
1074 out_bad:
1075 	ret = false;
1076 	goto out;
1077 }
1078 EXPORT_SYMBOL(__skb_flow_dissect);
1079 
1080 static u32 hashrnd __read_mostly;
1081 static __always_inline void __flow_hash_secret_init(void)
1082 {
1083 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1084 }
1085 
1086 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1087 					     u32 keyval)
1088 {
1089 	return jhash2(words, length, keyval);
1090 }
1091 
1092 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1093 {
1094 	const void *p = flow;
1095 
1096 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1097 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1098 }
1099 
1100 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1101 {
1102 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1103 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1104 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1105 		     sizeof(*flow) - sizeof(flow->addrs));
1106 
1107 	switch (flow->control.addr_type) {
1108 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1109 		diff -= sizeof(flow->addrs.v4addrs);
1110 		break;
1111 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1112 		diff -= sizeof(flow->addrs.v6addrs);
1113 		break;
1114 	case FLOW_DISSECTOR_KEY_TIPC:
1115 		diff -= sizeof(flow->addrs.tipckey);
1116 		break;
1117 	}
1118 	return (sizeof(*flow) - diff) / sizeof(u32);
1119 }
1120 
1121 __be32 flow_get_u32_src(const struct flow_keys *flow)
1122 {
1123 	switch (flow->control.addr_type) {
1124 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1125 		return flow->addrs.v4addrs.src;
1126 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1127 		return (__force __be32)ipv6_addr_hash(
1128 			&flow->addrs.v6addrs.src);
1129 	case FLOW_DISSECTOR_KEY_TIPC:
1130 		return flow->addrs.tipckey.key;
1131 	default:
1132 		return 0;
1133 	}
1134 }
1135 EXPORT_SYMBOL(flow_get_u32_src);
1136 
1137 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1138 {
1139 	switch (flow->control.addr_type) {
1140 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1141 		return flow->addrs.v4addrs.dst;
1142 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1143 		return (__force __be32)ipv6_addr_hash(
1144 			&flow->addrs.v6addrs.dst);
1145 	default:
1146 		return 0;
1147 	}
1148 }
1149 EXPORT_SYMBOL(flow_get_u32_dst);
1150 
1151 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1152 {
1153 	int addr_diff, i;
1154 
1155 	switch (keys->control.addr_type) {
1156 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1157 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1158 			    (__force u32)keys->addrs.v4addrs.src;
1159 		if ((addr_diff < 0) ||
1160 		    (addr_diff == 0 &&
1161 		     ((__force u16)keys->ports.dst <
1162 		      (__force u16)keys->ports.src))) {
1163 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1164 			swap(keys->ports.src, keys->ports.dst);
1165 		}
1166 		break;
1167 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1168 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1169 				   &keys->addrs.v6addrs.src,
1170 				   sizeof(keys->addrs.v6addrs.dst));
1171 		if ((addr_diff < 0) ||
1172 		    (addr_diff == 0 &&
1173 		     ((__force u16)keys->ports.dst <
1174 		      (__force u16)keys->ports.src))) {
1175 			for (i = 0; i < 4; i++)
1176 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1177 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1178 			swap(keys->ports.src, keys->ports.dst);
1179 		}
1180 		break;
1181 	}
1182 }
1183 
1184 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1185 {
1186 	u32 hash;
1187 
1188 	__flow_hash_consistentify(keys);
1189 
1190 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1191 				 flow_keys_hash_length(keys), keyval);
1192 	if (!hash)
1193 		hash = 1;
1194 
1195 	return hash;
1196 }
1197 
1198 u32 flow_hash_from_keys(struct flow_keys *keys)
1199 {
1200 	__flow_hash_secret_init();
1201 	return __flow_hash_from_keys(keys, hashrnd);
1202 }
1203 EXPORT_SYMBOL(flow_hash_from_keys);
1204 
1205 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1206 				  struct flow_keys *keys, u32 keyval)
1207 {
1208 	skb_flow_dissect_flow_keys(skb, keys,
1209 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1210 
1211 	return __flow_hash_from_keys(keys, keyval);
1212 }
1213 
1214 struct _flow_keys_digest_data {
1215 	__be16	n_proto;
1216 	u8	ip_proto;
1217 	u8	padding;
1218 	__be32	ports;
1219 	__be32	src;
1220 	__be32	dst;
1221 };
1222 
1223 void make_flow_keys_digest(struct flow_keys_digest *digest,
1224 			   const struct flow_keys *flow)
1225 {
1226 	struct _flow_keys_digest_data *data =
1227 	    (struct _flow_keys_digest_data *)digest;
1228 
1229 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1230 
1231 	memset(digest, 0, sizeof(*digest));
1232 
1233 	data->n_proto = flow->basic.n_proto;
1234 	data->ip_proto = flow->basic.ip_proto;
1235 	data->ports = flow->ports.ports;
1236 	data->src = flow->addrs.v4addrs.src;
1237 	data->dst = flow->addrs.v4addrs.dst;
1238 }
1239 EXPORT_SYMBOL(make_flow_keys_digest);
1240 
1241 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1242 
1243 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1244 {
1245 	struct flow_keys keys;
1246 
1247 	__flow_hash_secret_init();
1248 
1249 	memset(&keys, 0, sizeof(keys));
1250 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1251 			   NULL, 0, 0, 0,
1252 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1253 
1254 	return __flow_hash_from_keys(&keys, hashrnd);
1255 }
1256 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1257 
1258 /**
1259  * __skb_get_hash: calculate a flow hash
1260  * @skb: sk_buff to calculate flow hash from
1261  *
1262  * This function calculates a flow hash based on src/dst addresses
1263  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1264  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1265  * if hash is a canonical 4-tuple hash over transport ports.
1266  */
1267 void __skb_get_hash(struct sk_buff *skb)
1268 {
1269 	struct flow_keys keys;
1270 	u32 hash;
1271 
1272 	__flow_hash_secret_init();
1273 
1274 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1275 
1276 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1277 }
1278 EXPORT_SYMBOL(__skb_get_hash);
1279 
1280 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1281 {
1282 	struct flow_keys keys;
1283 
1284 	return ___skb_get_hash(skb, &keys, perturb);
1285 }
1286 EXPORT_SYMBOL(skb_get_hash_perturb);
1287 
1288 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1289 		   const struct flow_keys_basic *keys, int hlen)
1290 {
1291 	u32 poff = keys->control.thoff;
1292 
1293 	/* skip L4 headers for fragments after the first */
1294 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1295 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1296 		return poff;
1297 
1298 	switch (keys->basic.ip_proto) {
1299 	case IPPROTO_TCP: {
1300 		/* access doff as u8 to avoid unaligned access */
1301 		const u8 *doff;
1302 		u8 _doff;
1303 
1304 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1305 					    data, hlen, &_doff);
1306 		if (!doff)
1307 			return poff;
1308 
1309 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1310 		break;
1311 	}
1312 	case IPPROTO_UDP:
1313 	case IPPROTO_UDPLITE:
1314 		poff += sizeof(struct udphdr);
1315 		break;
1316 	/* For the rest, we do not really care about header
1317 	 * extensions at this point for now.
1318 	 */
1319 	case IPPROTO_ICMP:
1320 		poff += sizeof(struct icmphdr);
1321 		break;
1322 	case IPPROTO_ICMPV6:
1323 		poff += sizeof(struct icmp6hdr);
1324 		break;
1325 	case IPPROTO_IGMP:
1326 		poff += sizeof(struct igmphdr);
1327 		break;
1328 	case IPPROTO_DCCP:
1329 		poff += sizeof(struct dccp_hdr);
1330 		break;
1331 	case IPPROTO_SCTP:
1332 		poff += sizeof(struct sctphdr);
1333 		break;
1334 	}
1335 
1336 	return poff;
1337 }
1338 
1339 /**
1340  * skb_get_poff - get the offset to the payload
1341  * @skb: sk_buff to get the payload offset from
1342  *
1343  * The function will get the offset to the payload as far as it could
1344  * be dissected.  The main user is currently BPF, so that we can dynamically
1345  * truncate packets without needing to push actual payload to the user
1346  * space and can analyze headers only, instead.
1347  */
1348 u32 skb_get_poff(const struct sk_buff *skb)
1349 {
1350 	struct flow_keys_basic keys;
1351 
1352 	if (!skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0))
1353 		return 0;
1354 
1355 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1356 }
1357 
1358 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1359 {
1360 	memset(keys, 0, sizeof(*keys));
1361 
1362 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1363 	    sizeof(keys->addrs.v6addrs.src));
1364 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1365 	    sizeof(keys->addrs.v6addrs.dst));
1366 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1367 	keys->ports.src = fl6->fl6_sport;
1368 	keys->ports.dst = fl6->fl6_dport;
1369 	keys->keyid.keyid = fl6->fl6_gre_key;
1370 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1371 	keys->basic.ip_proto = fl6->flowi6_proto;
1372 
1373 	return flow_hash_from_keys(keys);
1374 }
1375 EXPORT_SYMBOL(__get_hash_from_flowi6);
1376 
1377 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1378 	{
1379 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1380 		.offset = offsetof(struct flow_keys, control),
1381 	},
1382 	{
1383 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1384 		.offset = offsetof(struct flow_keys, basic),
1385 	},
1386 	{
1387 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1388 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1389 	},
1390 	{
1391 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1392 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1393 	},
1394 	{
1395 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1396 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1397 	},
1398 	{
1399 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1400 		.offset = offsetof(struct flow_keys, ports),
1401 	},
1402 	{
1403 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1404 		.offset = offsetof(struct flow_keys, vlan),
1405 	},
1406 	{
1407 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1408 		.offset = offsetof(struct flow_keys, tags),
1409 	},
1410 	{
1411 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1412 		.offset = offsetof(struct flow_keys, keyid),
1413 	},
1414 };
1415 
1416 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1417 	{
1418 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1419 		.offset = offsetof(struct flow_keys, control),
1420 	},
1421 	{
1422 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1423 		.offset = offsetof(struct flow_keys, basic),
1424 	},
1425 	{
1426 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1427 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1428 	},
1429 	{
1430 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1431 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1432 	},
1433 	{
1434 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1435 		.offset = offsetof(struct flow_keys, ports),
1436 	},
1437 };
1438 
1439 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1440 	{
1441 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1442 		.offset = offsetof(struct flow_keys, control),
1443 	},
1444 	{
1445 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1446 		.offset = offsetof(struct flow_keys, basic),
1447 	},
1448 };
1449 
1450 struct flow_dissector flow_keys_dissector __read_mostly;
1451 EXPORT_SYMBOL(flow_keys_dissector);
1452 
1453 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1454 EXPORT_SYMBOL(flow_keys_basic_dissector);
1455 
1456 static int __init init_default_flow_dissectors(void)
1457 {
1458 	skb_flow_dissector_init(&flow_keys_dissector,
1459 				flow_keys_dissector_keys,
1460 				ARRAY_SIZE(flow_keys_dissector_keys));
1461 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1462 				flow_keys_dissector_symmetric_keys,
1463 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1464 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1465 				flow_keys_basic_dissector_keys,
1466 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1467 	return 0;
1468 }
1469 
1470 core_initcall(init_default_flow_dissectors);
1471