xref: /openbmc/linux/net/core/flow_dissector.c (revision a17922de)
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 		return;
159 
160 	info = skb_tunnel_info(skb);
161 	if (!info)
162 		return;
163 
164 	key = &info->key;
165 
166 	switch (ip_tunnel_info_af(info)) {
167 	case AF_INET:
168 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
169 						   flow_dissector,
170 						   target_container);
171 		if (dissector_uses_key(flow_dissector,
172 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
173 			struct flow_dissector_key_ipv4_addrs *ipv4;
174 
175 			ipv4 = skb_flow_dissector_target(flow_dissector,
176 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
177 							 target_container);
178 			ipv4->src = key->u.ipv4.src;
179 			ipv4->dst = key->u.ipv4.dst;
180 		}
181 		break;
182 	case AF_INET6:
183 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
184 						   flow_dissector,
185 						   target_container);
186 		if (dissector_uses_key(flow_dissector,
187 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
188 			struct flow_dissector_key_ipv6_addrs *ipv6;
189 
190 			ipv6 = skb_flow_dissector_target(flow_dissector,
191 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
192 							 target_container);
193 			ipv6->src = key->u.ipv6.src;
194 			ipv6->dst = key->u.ipv6.dst;
195 		}
196 		break;
197 	}
198 
199 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
200 		struct flow_dissector_key_keyid *keyid;
201 
202 		keyid = skb_flow_dissector_target(flow_dissector,
203 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
204 						  target_container);
205 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
206 	}
207 
208 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
209 		struct flow_dissector_key_ports *tp;
210 
211 		tp = skb_flow_dissector_target(flow_dissector,
212 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
213 					       target_container);
214 		tp->src = key->tp_src;
215 		tp->dst = key->tp_dst;
216 	}
217 
218 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
219 		struct flow_dissector_key_ip *ip;
220 
221 		ip = skb_flow_dissector_target(flow_dissector,
222 					       FLOW_DISSECTOR_KEY_ENC_IP,
223 					       target_container);
224 		ip->tos = key->tos;
225 		ip->ttl = key->ttl;
226 	}
227 }
228 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
229 
230 static enum flow_dissect_ret
231 __skb_flow_dissect_mpls(const struct sk_buff *skb,
232 			struct flow_dissector *flow_dissector,
233 			void *target_container, void *data, int nhoff, int hlen)
234 {
235 	struct flow_dissector_key_keyid *key_keyid;
236 	struct mpls_label *hdr, _hdr[2];
237 	u32 entry, label;
238 
239 	if (!dissector_uses_key(flow_dissector,
240 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
241 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
242 		return FLOW_DISSECT_RET_OUT_GOOD;
243 
244 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
245 				   hlen, &_hdr);
246 	if (!hdr)
247 		return FLOW_DISSECT_RET_OUT_BAD;
248 
249 	entry = ntohl(hdr[0].entry);
250 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
251 
252 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
253 		struct flow_dissector_key_mpls *key_mpls;
254 
255 		key_mpls = skb_flow_dissector_target(flow_dissector,
256 						     FLOW_DISSECTOR_KEY_MPLS,
257 						     target_container);
258 		key_mpls->mpls_label = label;
259 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
260 					>> MPLS_LS_TTL_SHIFT;
261 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
262 					>> MPLS_LS_TC_SHIFT;
263 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
264 					>> MPLS_LS_S_SHIFT;
265 	}
266 
267 	if (label == MPLS_LABEL_ENTROPY) {
268 		key_keyid = skb_flow_dissector_target(flow_dissector,
269 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
270 						      target_container);
271 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
272 	}
273 	return FLOW_DISSECT_RET_OUT_GOOD;
274 }
275 
276 static enum flow_dissect_ret
277 __skb_flow_dissect_arp(const struct sk_buff *skb,
278 		       struct flow_dissector *flow_dissector,
279 		       void *target_container, void *data, int nhoff, int hlen)
280 {
281 	struct flow_dissector_key_arp *key_arp;
282 	struct {
283 		unsigned char ar_sha[ETH_ALEN];
284 		unsigned char ar_sip[4];
285 		unsigned char ar_tha[ETH_ALEN];
286 		unsigned char ar_tip[4];
287 	} *arp_eth, _arp_eth;
288 	const struct arphdr *arp;
289 	struct arphdr _arp;
290 
291 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
292 		return FLOW_DISSECT_RET_OUT_GOOD;
293 
294 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
295 				   hlen, &_arp);
296 	if (!arp)
297 		return FLOW_DISSECT_RET_OUT_BAD;
298 
299 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
300 	    arp->ar_pro != htons(ETH_P_IP) ||
301 	    arp->ar_hln != ETH_ALEN ||
302 	    arp->ar_pln != 4 ||
303 	    (arp->ar_op != htons(ARPOP_REPLY) &&
304 	     arp->ar_op != htons(ARPOP_REQUEST)))
305 		return FLOW_DISSECT_RET_OUT_BAD;
306 
307 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
308 				       sizeof(_arp_eth), data,
309 				       hlen, &_arp_eth);
310 	if (!arp_eth)
311 		return FLOW_DISSECT_RET_OUT_BAD;
312 
313 	key_arp = skb_flow_dissector_target(flow_dissector,
314 					    FLOW_DISSECTOR_KEY_ARP,
315 					    target_container);
316 
317 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
318 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
319 
320 	/* Only store the lower byte of the opcode;
321 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
322 	 */
323 	key_arp->op = ntohs(arp->ar_op) & 0xff;
324 
325 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
326 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
327 
328 	return FLOW_DISSECT_RET_OUT_GOOD;
329 }
330 
331 static enum flow_dissect_ret
332 __skb_flow_dissect_gre(const struct sk_buff *skb,
333 		       struct flow_dissector_key_control *key_control,
334 		       struct flow_dissector *flow_dissector,
335 		       void *target_container, void *data,
336 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
337 		       unsigned int flags)
338 {
339 	struct flow_dissector_key_keyid *key_keyid;
340 	struct gre_base_hdr *hdr, _hdr;
341 	int offset = 0;
342 	u16 gre_ver;
343 
344 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
345 				   data, *p_hlen, &_hdr);
346 	if (!hdr)
347 		return FLOW_DISSECT_RET_OUT_BAD;
348 
349 	/* Only look inside GRE without routing */
350 	if (hdr->flags & GRE_ROUTING)
351 		return FLOW_DISSECT_RET_OUT_GOOD;
352 
353 	/* Only look inside GRE for version 0 and 1 */
354 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
355 	if (gre_ver > 1)
356 		return FLOW_DISSECT_RET_OUT_GOOD;
357 
358 	*p_proto = hdr->protocol;
359 	if (gre_ver) {
360 		/* Version1 must be PPTP, and check the flags */
361 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
362 			return FLOW_DISSECT_RET_OUT_GOOD;
363 	}
364 
365 	offset += sizeof(struct gre_base_hdr);
366 
367 	if (hdr->flags & GRE_CSUM)
368 		offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
369 			  sizeof(((struct gre_full_hdr *) 0)->reserved1);
370 
371 	if (hdr->flags & GRE_KEY) {
372 		const __be32 *keyid;
373 		__be32 _keyid;
374 
375 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
376 					     sizeof(_keyid),
377 					     data, *p_hlen, &_keyid);
378 		if (!keyid)
379 			return FLOW_DISSECT_RET_OUT_BAD;
380 
381 		if (dissector_uses_key(flow_dissector,
382 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
383 			key_keyid = skb_flow_dissector_target(flow_dissector,
384 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
385 							      target_container);
386 			if (gre_ver == 0)
387 				key_keyid->keyid = *keyid;
388 			else
389 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
390 		}
391 		offset += sizeof(((struct gre_full_hdr *) 0)->key);
392 	}
393 
394 	if (hdr->flags & GRE_SEQ)
395 		offset += sizeof(((struct pptp_gre_header *) 0)->seq);
396 
397 	if (gre_ver == 0) {
398 		if (*p_proto == htons(ETH_P_TEB)) {
399 			const struct ethhdr *eth;
400 			struct ethhdr _eth;
401 
402 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
403 						   sizeof(_eth),
404 						   data, *p_hlen, &_eth);
405 			if (!eth)
406 				return FLOW_DISSECT_RET_OUT_BAD;
407 			*p_proto = eth->h_proto;
408 			offset += sizeof(*eth);
409 
410 			/* Cap headers that we access via pointers at the
411 			 * end of the Ethernet header as our maximum alignment
412 			 * at that point is only 2 bytes.
413 			 */
414 			if (NET_IP_ALIGN)
415 				*p_hlen = *p_nhoff + offset;
416 		}
417 	} else { /* version 1, must be PPTP */
418 		u8 _ppp_hdr[PPP_HDRLEN];
419 		u8 *ppp_hdr;
420 
421 		if (hdr->flags & GRE_ACK)
422 			offset += sizeof(((struct pptp_gre_header *) 0)->ack);
423 
424 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
425 					       sizeof(_ppp_hdr),
426 					       data, *p_hlen, _ppp_hdr);
427 		if (!ppp_hdr)
428 			return FLOW_DISSECT_RET_OUT_BAD;
429 
430 		switch (PPP_PROTOCOL(ppp_hdr)) {
431 		case PPP_IP:
432 			*p_proto = htons(ETH_P_IP);
433 			break;
434 		case PPP_IPV6:
435 			*p_proto = htons(ETH_P_IPV6);
436 			break;
437 		default:
438 			/* Could probably catch some more like MPLS */
439 			break;
440 		}
441 
442 		offset += PPP_HDRLEN;
443 	}
444 
445 	*p_nhoff += offset;
446 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
447 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
448 		return FLOW_DISSECT_RET_OUT_GOOD;
449 
450 	return FLOW_DISSECT_RET_PROTO_AGAIN;
451 }
452 
453 /**
454  * __skb_flow_dissect_batadv() - dissect batman-adv header
455  * @skb: sk_buff to with the batman-adv header
456  * @key_control: flow dissectors control key
457  * @data: raw buffer pointer to the packet, if NULL use skb->data
458  * @p_proto: pointer used to update the protocol to process next
459  * @p_nhoff: pointer used to update inner network header offset
460  * @hlen: packet header length
461  * @flags: any combination of FLOW_DISSECTOR_F_*
462  *
463  * ETH_P_BATMAN packets are tried to be dissected. Only
464  * &struct batadv_unicast packets are actually processed because they contain an
465  * inner ethernet header and are usually followed by actual network header. This
466  * allows the flow dissector to continue processing the packet.
467  *
468  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
469  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
470  *  otherwise FLOW_DISSECT_RET_OUT_BAD
471  */
472 static enum flow_dissect_ret
473 __skb_flow_dissect_batadv(const struct sk_buff *skb,
474 			  struct flow_dissector_key_control *key_control,
475 			  void *data, __be16 *p_proto, int *p_nhoff, int hlen,
476 			  unsigned int flags)
477 {
478 	struct {
479 		struct batadv_unicast_packet batadv_unicast;
480 		struct ethhdr eth;
481 	} *hdr, _hdr;
482 
483 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
484 				   &_hdr);
485 	if (!hdr)
486 		return FLOW_DISSECT_RET_OUT_BAD;
487 
488 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
489 		return FLOW_DISSECT_RET_OUT_BAD;
490 
491 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
492 		return FLOW_DISSECT_RET_OUT_BAD;
493 
494 	*p_proto = hdr->eth.h_proto;
495 	*p_nhoff += sizeof(*hdr);
496 
497 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
498 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
499 		return FLOW_DISSECT_RET_OUT_GOOD;
500 
501 	return FLOW_DISSECT_RET_PROTO_AGAIN;
502 }
503 
504 static void
505 __skb_flow_dissect_tcp(const struct sk_buff *skb,
506 		       struct flow_dissector *flow_dissector,
507 		       void *target_container, void *data, int thoff, int hlen)
508 {
509 	struct flow_dissector_key_tcp *key_tcp;
510 	struct tcphdr *th, _th;
511 
512 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
513 		return;
514 
515 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
516 	if (!th)
517 		return;
518 
519 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
520 		return;
521 
522 	key_tcp = skb_flow_dissector_target(flow_dissector,
523 					    FLOW_DISSECTOR_KEY_TCP,
524 					    target_container);
525 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
526 }
527 
528 static void
529 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
530 			struct flow_dissector *flow_dissector,
531 			void *target_container, void *data, const struct iphdr *iph)
532 {
533 	struct flow_dissector_key_ip *key_ip;
534 
535 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
536 		return;
537 
538 	key_ip = skb_flow_dissector_target(flow_dissector,
539 					   FLOW_DISSECTOR_KEY_IP,
540 					   target_container);
541 	key_ip->tos = iph->tos;
542 	key_ip->ttl = iph->ttl;
543 }
544 
545 static void
546 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
547 			struct flow_dissector *flow_dissector,
548 			void *target_container, void *data, const struct ipv6hdr *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 = ipv6_get_dsfield(iph);
559 	key_ip->ttl = iph->hop_limit;
560 }
561 
562 /* Maximum number of protocol headers that can be parsed in
563  * __skb_flow_dissect
564  */
565 #define MAX_FLOW_DISSECT_HDRS	15
566 
567 static bool skb_flow_dissect_allowed(int *num_hdrs)
568 {
569 	++*num_hdrs;
570 
571 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
572 }
573 
574 /**
575  * __skb_flow_dissect - extract the flow_keys struct and return it
576  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
577  * @flow_dissector: list of keys to dissect
578  * @target_container: target structure to put dissected values into
579  * @data: raw buffer pointer to the packet, if NULL use skb->data
580  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
581  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
582  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
583  *
584  * The function will try to retrieve individual keys into target specified
585  * by flow_dissector from either the skbuff or a raw buffer specified by the
586  * rest parameters.
587  *
588  * Caller must take care of zeroing target container memory.
589  */
590 bool __skb_flow_dissect(const struct sk_buff *skb,
591 			struct flow_dissector *flow_dissector,
592 			void *target_container,
593 			void *data, __be16 proto, int nhoff, int hlen,
594 			unsigned int flags)
595 {
596 	struct flow_dissector_key_control *key_control;
597 	struct flow_dissector_key_basic *key_basic;
598 	struct flow_dissector_key_addrs *key_addrs;
599 	struct flow_dissector_key_ports *key_ports;
600 	struct flow_dissector_key_icmp *key_icmp;
601 	struct flow_dissector_key_tags *key_tags;
602 	struct flow_dissector_key_vlan *key_vlan;
603 	enum flow_dissect_ret fdret;
604 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
605 	int num_hdrs = 0;
606 	u8 ip_proto = 0;
607 	bool ret;
608 
609 	if (!data) {
610 		data = skb->data;
611 		proto = skb_vlan_tag_present(skb) ?
612 			 skb->vlan_proto : skb->protocol;
613 		nhoff = skb_network_offset(skb);
614 		hlen = skb_headlen(skb);
615 #if IS_ENABLED(CONFIG_NET_DSA)
616 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
617 			const struct dsa_device_ops *ops;
618 			int offset;
619 
620 			ops = skb->dev->dsa_ptr->tag_ops;
621 			if (ops->flow_dissect &&
622 			    !ops->flow_dissect(skb, &proto, &offset)) {
623 				hlen -= offset;
624 				nhoff += offset;
625 			}
626 		}
627 #endif
628 	}
629 
630 	/* It is ensured by skb_flow_dissector_init() that control key will
631 	 * be always present.
632 	 */
633 	key_control = skb_flow_dissector_target(flow_dissector,
634 						FLOW_DISSECTOR_KEY_CONTROL,
635 						target_container);
636 
637 	/* It is ensured by skb_flow_dissector_init() that basic key will
638 	 * be always present.
639 	 */
640 	key_basic = skb_flow_dissector_target(flow_dissector,
641 					      FLOW_DISSECTOR_KEY_BASIC,
642 					      target_container);
643 
644 	if (dissector_uses_key(flow_dissector,
645 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
646 		struct ethhdr *eth = eth_hdr(skb);
647 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
648 
649 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
650 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
651 							  target_container);
652 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
653 	}
654 
655 proto_again:
656 	fdret = FLOW_DISSECT_RET_CONTINUE;
657 
658 	switch (proto) {
659 	case htons(ETH_P_IP): {
660 		const struct iphdr *iph;
661 		struct iphdr _iph;
662 
663 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
664 		if (!iph || iph->ihl < 5) {
665 			fdret = FLOW_DISSECT_RET_OUT_BAD;
666 			break;
667 		}
668 
669 		nhoff += iph->ihl * 4;
670 
671 		ip_proto = iph->protocol;
672 
673 		if (dissector_uses_key(flow_dissector,
674 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
675 			key_addrs = skb_flow_dissector_target(flow_dissector,
676 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
677 							      target_container);
678 
679 			memcpy(&key_addrs->v4addrs, &iph->saddr,
680 			       sizeof(key_addrs->v4addrs));
681 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
682 		}
683 
684 		if (ip_is_fragment(iph)) {
685 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
686 
687 			if (iph->frag_off & htons(IP_OFFSET)) {
688 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
689 				break;
690 			} else {
691 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
692 				if (!(flags &
693 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
694 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
695 					break;
696 				}
697 			}
698 		}
699 
700 		__skb_flow_dissect_ipv4(skb, flow_dissector,
701 					target_container, data, iph);
702 
703 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
704 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
705 			break;
706 		}
707 
708 		break;
709 	}
710 	case htons(ETH_P_IPV6): {
711 		const struct ipv6hdr *iph;
712 		struct ipv6hdr _iph;
713 
714 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
715 		if (!iph) {
716 			fdret = FLOW_DISSECT_RET_OUT_BAD;
717 			break;
718 		}
719 
720 		ip_proto = iph->nexthdr;
721 		nhoff += sizeof(struct ipv6hdr);
722 
723 		if (dissector_uses_key(flow_dissector,
724 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
725 			key_addrs = skb_flow_dissector_target(flow_dissector,
726 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
727 							      target_container);
728 
729 			memcpy(&key_addrs->v6addrs, &iph->saddr,
730 			       sizeof(key_addrs->v6addrs));
731 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
732 		}
733 
734 		if ((dissector_uses_key(flow_dissector,
735 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
736 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
737 		    ip6_flowlabel(iph)) {
738 			__be32 flow_label = ip6_flowlabel(iph);
739 
740 			if (dissector_uses_key(flow_dissector,
741 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
742 				key_tags = skb_flow_dissector_target(flow_dissector,
743 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
744 								     target_container);
745 				key_tags->flow_label = ntohl(flow_label);
746 			}
747 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
748 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
749 				break;
750 			}
751 		}
752 
753 		__skb_flow_dissect_ipv6(skb, flow_dissector,
754 					target_container, data, iph);
755 
756 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
757 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
758 
759 		break;
760 	}
761 	case htons(ETH_P_8021AD):
762 	case htons(ETH_P_8021Q): {
763 		const struct vlan_hdr *vlan = NULL;
764 		struct vlan_hdr _vlan;
765 		__be16 saved_vlan_tpid = proto;
766 
767 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
768 		    skb && skb_vlan_tag_present(skb)) {
769 			proto = skb->protocol;
770 		} else {
771 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
772 						    data, hlen, &_vlan);
773 			if (!vlan) {
774 				fdret = FLOW_DISSECT_RET_OUT_BAD;
775 				break;
776 			}
777 
778 			proto = vlan->h_vlan_encapsulated_proto;
779 			nhoff += sizeof(*vlan);
780 		}
781 
782 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
783 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
784 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
785 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
786 		} else {
787 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
788 			break;
789 		}
790 
791 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
792 			key_vlan = skb_flow_dissector_target(flow_dissector,
793 							     dissector_vlan,
794 							     target_container);
795 
796 			if (!vlan) {
797 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
798 				key_vlan->vlan_priority =
799 					(skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
800 			} else {
801 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
802 					VLAN_VID_MASK;
803 				key_vlan->vlan_priority =
804 					(ntohs(vlan->h_vlan_TCI) &
805 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
806 			}
807 			key_vlan->vlan_tpid = saved_vlan_tpid;
808 		}
809 
810 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
811 		break;
812 	}
813 	case htons(ETH_P_PPP_SES): {
814 		struct {
815 			struct pppoe_hdr hdr;
816 			__be16 proto;
817 		} *hdr, _hdr;
818 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
819 		if (!hdr) {
820 			fdret = FLOW_DISSECT_RET_OUT_BAD;
821 			break;
822 		}
823 
824 		proto = hdr->proto;
825 		nhoff += PPPOE_SES_HLEN;
826 		switch (proto) {
827 		case htons(PPP_IP):
828 			proto = htons(ETH_P_IP);
829 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
830 			break;
831 		case htons(PPP_IPV6):
832 			proto = htons(ETH_P_IPV6);
833 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
834 			break;
835 		default:
836 			fdret = FLOW_DISSECT_RET_OUT_BAD;
837 			break;
838 		}
839 		break;
840 	}
841 	case htons(ETH_P_TIPC): {
842 		struct tipc_basic_hdr *hdr, _hdr;
843 
844 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
845 					   data, hlen, &_hdr);
846 		if (!hdr) {
847 			fdret = FLOW_DISSECT_RET_OUT_BAD;
848 			break;
849 		}
850 
851 		if (dissector_uses_key(flow_dissector,
852 				       FLOW_DISSECTOR_KEY_TIPC)) {
853 			key_addrs = skb_flow_dissector_target(flow_dissector,
854 							      FLOW_DISSECTOR_KEY_TIPC,
855 							      target_container);
856 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
857 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
858 		}
859 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
860 		break;
861 	}
862 
863 	case htons(ETH_P_MPLS_UC):
864 	case htons(ETH_P_MPLS_MC):
865 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
866 						target_container, data,
867 						nhoff, hlen);
868 		break;
869 	case htons(ETH_P_FCOE):
870 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
871 			fdret = FLOW_DISSECT_RET_OUT_BAD;
872 			break;
873 		}
874 
875 		nhoff += FCOE_HEADER_LEN;
876 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
877 		break;
878 
879 	case htons(ETH_P_ARP):
880 	case htons(ETH_P_RARP):
881 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
882 					       target_container, data,
883 					       nhoff, hlen);
884 		break;
885 
886 	case htons(ETH_P_BATMAN):
887 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
888 						  &proto, &nhoff, hlen, flags);
889 		break;
890 
891 	default:
892 		fdret = FLOW_DISSECT_RET_OUT_BAD;
893 		break;
894 	}
895 
896 	/* Process result of proto processing */
897 	switch (fdret) {
898 	case FLOW_DISSECT_RET_OUT_GOOD:
899 		goto out_good;
900 	case FLOW_DISSECT_RET_PROTO_AGAIN:
901 		if (skb_flow_dissect_allowed(&num_hdrs))
902 			goto proto_again;
903 		goto out_good;
904 	case FLOW_DISSECT_RET_CONTINUE:
905 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
906 		break;
907 	case FLOW_DISSECT_RET_OUT_BAD:
908 	default:
909 		goto out_bad;
910 	}
911 
912 ip_proto_again:
913 	fdret = FLOW_DISSECT_RET_CONTINUE;
914 
915 	switch (ip_proto) {
916 	case IPPROTO_GRE:
917 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
918 					       target_container, data,
919 					       &proto, &nhoff, &hlen, flags);
920 		break;
921 
922 	case NEXTHDR_HOP:
923 	case NEXTHDR_ROUTING:
924 	case NEXTHDR_DEST: {
925 		u8 _opthdr[2], *opthdr;
926 
927 		if (proto != htons(ETH_P_IPV6))
928 			break;
929 
930 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
931 					      data, hlen, &_opthdr);
932 		if (!opthdr) {
933 			fdret = FLOW_DISSECT_RET_OUT_BAD;
934 			break;
935 		}
936 
937 		ip_proto = opthdr[0];
938 		nhoff += (opthdr[1] + 1) << 3;
939 
940 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
941 		break;
942 	}
943 	case NEXTHDR_FRAGMENT: {
944 		struct frag_hdr _fh, *fh;
945 
946 		if (proto != htons(ETH_P_IPV6))
947 			break;
948 
949 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
950 					  data, hlen, &_fh);
951 
952 		if (!fh) {
953 			fdret = FLOW_DISSECT_RET_OUT_BAD;
954 			break;
955 		}
956 
957 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
958 
959 		nhoff += sizeof(_fh);
960 		ip_proto = fh->nexthdr;
961 
962 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
963 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
964 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
965 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
966 				break;
967 			}
968 		}
969 
970 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
971 		break;
972 	}
973 	case IPPROTO_IPIP:
974 		proto = htons(ETH_P_IP);
975 
976 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
977 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
978 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
979 			break;
980 		}
981 
982 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
983 		break;
984 
985 	case IPPROTO_IPV6:
986 		proto = htons(ETH_P_IPV6);
987 
988 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
989 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
990 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
991 			break;
992 		}
993 
994 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
995 		break;
996 
997 
998 	case IPPROTO_MPLS:
999 		proto = htons(ETH_P_MPLS_UC);
1000 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1001 		break;
1002 
1003 	case IPPROTO_TCP:
1004 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1005 				       data, nhoff, hlen);
1006 		break;
1007 
1008 	default:
1009 		break;
1010 	}
1011 
1012 	if (dissector_uses_key(flow_dissector,
1013 			       FLOW_DISSECTOR_KEY_PORTS)) {
1014 		key_ports = skb_flow_dissector_target(flow_dissector,
1015 						      FLOW_DISSECTOR_KEY_PORTS,
1016 						      target_container);
1017 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1018 							data, hlen);
1019 	}
1020 
1021 	if (dissector_uses_key(flow_dissector,
1022 			       FLOW_DISSECTOR_KEY_ICMP)) {
1023 		key_icmp = skb_flow_dissector_target(flow_dissector,
1024 						     FLOW_DISSECTOR_KEY_ICMP,
1025 						     target_container);
1026 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1027 	}
1028 
1029 	/* Process result of IP proto processing */
1030 	switch (fdret) {
1031 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1032 		if (skb_flow_dissect_allowed(&num_hdrs))
1033 			goto proto_again;
1034 		break;
1035 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1036 		if (skb_flow_dissect_allowed(&num_hdrs))
1037 			goto ip_proto_again;
1038 		break;
1039 	case FLOW_DISSECT_RET_OUT_GOOD:
1040 	case FLOW_DISSECT_RET_CONTINUE:
1041 		break;
1042 	case FLOW_DISSECT_RET_OUT_BAD:
1043 	default:
1044 		goto out_bad;
1045 	}
1046 
1047 out_good:
1048 	ret = true;
1049 
1050 out:
1051 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1052 	key_basic->n_proto = proto;
1053 	key_basic->ip_proto = ip_proto;
1054 
1055 	return ret;
1056 
1057 out_bad:
1058 	ret = false;
1059 	goto out;
1060 }
1061 EXPORT_SYMBOL(__skb_flow_dissect);
1062 
1063 static u32 hashrnd __read_mostly;
1064 static __always_inline void __flow_hash_secret_init(void)
1065 {
1066 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1067 }
1068 
1069 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1070 					     u32 keyval)
1071 {
1072 	return jhash2(words, length, keyval);
1073 }
1074 
1075 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1076 {
1077 	const void *p = flow;
1078 
1079 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1080 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1081 }
1082 
1083 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1084 {
1085 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1086 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1087 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1088 		     sizeof(*flow) - sizeof(flow->addrs));
1089 
1090 	switch (flow->control.addr_type) {
1091 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1092 		diff -= sizeof(flow->addrs.v4addrs);
1093 		break;
1094 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1095 		diff -= sizeof(flow->addrs.v6addrs);
1096 		break;
1097 	case FLOW_DISSECTOR_KEY_TIPC:
1098 		diff -= sizeof(flow->addrs.tipckey);
1099 		break;
1100 	}
1101 	return (sizeof(*flow) - diff) / sizeof(u32);
1102 }
1103 
1104 __be32 flow_get_u32_src(const struct flow_keys *flow)
1105 {
1106 	switch (flow->control.addr_type) {
1107 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1108 		return flow->addrs.v4addrs.src;
1109 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1110 		return (__force __be32)ipv6_addr_hash(
1111 			&flow->addrs.v6addrs.src);
1112 	case FLOW_DISSECTOR_KEY_TIPC:
1113 		return flow->addrs.tipckey.key;
1114 	default:
1115 		return 0;
1116 	}
1117 }
1118 EXPORT_SYMBOL(flow_get_u32_src);
1119 
1120 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1121 {
1122 	switch (flow->control.addr_type) {
1123 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1124 		return flow->addrs.v4addrs.dst;
1125 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1126 		return (__force __be32)ipv6_addr_hash(
1127 			&flow->addrs.v6addrs.dst);
1128 	default:
1129 		return 0;
1130 	}
1131 }
1132 EXPORT_SYMBOL(flow_get_u32_dst);
1133 
1134 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1135 {
1136 	int addr_diff, i;
1137 
1138 	switch (keys->control.addr_type) {
1139 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1140 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1141 			    (__force u32)keys->addrs.v4addrs.src;
1142 		if ((addr_diff < 0) ||
1143 		    (addr_diff == 0 &&
1144 		     ((__force u16)keys->ports.dst <
1145 		      (__force u16)keys->ports.src))) {
1146 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1147 			swap(keys->ports.src, keys->ports.dst);
1148 		}
1149 		break;
1150 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1151 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1152 				   &keys->addrs.v6addrs.src,
1153 				   sizeof(keys->addrs.v6addrs.dst));
1154 		if ((addr_diff < 0) ||
1155 		    (addr_diff == 0 &&
1156 		     ((__force u16)keys->ports.dst <
1157 		      (__force u16)keys->ports.src))) {
1158 			for (i = 0; i < 4; i++)
1159 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1160 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1161 			swap(keys->ports.src, keys->ports.dst);
1162 		}
1163 		break;
1164 	}
1165 }
1166 
1167 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1168 {
1169 	u32 hash;
1170 
1171 	__flow_hash_consistentify(keys);
1172 
1173 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1174 				 flow_keys_hash_length(keys), keyval);
1175 	if (!hash)
1176 		hash = 1;
1177 
1178 	return hash;
1179 }
1180 
1181 u32 flow_hash_from_keys(struct flow_keys *keys)
1182 {
1183 	__flow_hash_secret_init();
1184 	return __flow_hash_from_keys(keys, hashrnd);
1185 }
1186 EXPORT_SYMBOL(flow_hash_from_keys);
1187 
1188 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1189 				  struct flow_keys *keys, u32 keyval)
1190 {
1191 	skb_flow_dissect_flow_keys(skb, keys,
1192 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1193 
1194 	return __flow_hash_from_keys(keys, keyval);
1195 }
1196 
1197 struct _flow_keys_digest_data {
1198 	__be16	n_proto;
1199 	u8	ip_proto;
1200 	u8	padding;
1201 	__be32	ports;
1202 	__be32	src;
1203 	__be32	dst;
1204 };
1205 
1206 void make_flow_keys_digest(struct flow_keys_digest *digest,
1207 			   const struct flow_keys *flow)
1208 {
1209 	struct _flow_keys_digest_data *data =
1210 	    (struct _flow_keys_digest_data *)digest;
1211 
1212 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1213 
1214 	memset(digest, 0, sizeof(*digest));
1215 
1216 	data->n_proto = flow->basic.n_proto;
1217 	data->ip_proto = flow->basic.ip_proto;
1218 	data->ports = flow->ports.ports;
1219 	data->src = flow->addrs.v4addrs.src;
1220 	data->dst = flow->addrs.v4addrs.dst;
1221 }
1222 EXPORT_SYMBOL(make_flow_keys_digest);
1223 
1224 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1225 
1226 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1227 {
1228 	struct flow_keys keys;
1229 
1230 	__flow_hash_secret_init();
1231 
1232 	memset(&keys, 0, sizeof(keys));
1233 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1234 			   NULL, 0, 0, 0,
1235 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1236 
1237 	return __flow_hash_from_keys(&keys, hashrnd);
1238 }
1239 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1240 
1241 /**
1242  * __skb_get_hash: calculate a flow hash
1243  * @skb: sk_buff to calculate flow hash from
1244  *
1245  * This function calculates a flow hash based on src/dst addresses
1246  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1247  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1248  * if hash is a canonical 4-tuple hash over transport ports.
1249  */
1250 void __skb_get_hash(struct sk_buff *skb)
1251 {
1252 	struct flow_keys keys;
1253 	u32 hash;
1254 
1255 	__flow_hash_secret_init();
1256 
1257 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1258 
1259 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1260 }
1261 EXPORT_SYMBOL(__skb_get_hash);
1262 
1263 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1264 {
1265 	struct flow_keys keys;
1266 
1267 	return ___skb_get_hash(skb, &keys, perturb);
1268 }
1269 EXPORT_SYMBOL(skb_get_hash_perturb);
1270 
1271 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1272 		   const struct flow_keys_basic *keys, int hlen)
1273 {
1274 	u32 poff = keys->control.thoff;
1275 
1276 	/* skip L4 headers for fragments after the first */
1277 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1278 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1279 		return poff;
1280 
1281 	switch (keys->basic.ip_proto) {
1282 	case IPPROTO_TCP: {
1283 		/* access doff as u8 to avoid unaligned access */
1284 		const u8 *doff;
1285 		u8 _doff;
1286 
1287 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1288 					    data, hlen, &_doff);
1289 		if (!doff)
1290 			return poff;
1291 
1292 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1293 		break;
1294 	}
1295 	case IPPROTO_UDP:
1296 	case IPPROTO_UDPLITE:
1297 		poff += sizeof(struct udphdr);
1298 		break;
1299 	/* For the rest, we do not really care about header
1300 	 * extensions at this point for now.
1301 	 */
1302 	case IPPROTO_ICMP:
1303 		poff += sizeof(struct icmphdr);
1304 		break;
1305 	case IPPROTO_ICMPV6:
1306 		poff += sizeof(struct icmp6hdr);
1307 		break;
1308 	case IPPROTO_IGMP:
1309 		poff += sizeof(struct igmphdr);
1310 		break;
1311 	case IPPROTO_DCCP:
1312 		poff += sizeof(struct dccp_hdr);
1313 		break;
1314 	case IPPROTO_SCTP:
1315 		poff += sizeof(struct sctphdr);
1316 		break;
1317 	}
1318 
1319 	return poff;
1320 }
1321 
1322 /**
1323  * skb_get_poff - get the offset to the payload
1324  * @skb: sk_buff to get the payload offset from
1325  *
1326  * The function will get the offset to the payload as far as it could
1327  * be dissected.  The main user is currently BPF, so that we can dynamically
1328  * truncate packets without needing to push actual payload to the user
1329  * space and can analyze headers only, instead.
1330  */
1331 u32 skb_get_poff(const struct sk_buff *skb)
1332 {
1333 	struct flow_keys_basic keys;
1334 
1335 	if (!skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0))
1336 		return 0;
1337 
1338 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1339 }
1340 
1341 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1342 {
1343 	memset(keys, 0, sizeof(*keys));
1344 
1345 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1346 	    sizeof(keys->addrs.v6addrs.src));
1347 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1348 	    sizeof(keys->addrs.v6addrs.dst));
1349 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1350 	keys->ports.src = fl6->fl6_sport;
1351 	keys->ports.dst = fl6->fl6_dport;
1352 	keys->keyid.keyid = fl6->fl6_gre_key;
1353 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1354 	keys->basic.ip_proto = fl6->flowi6_proto;
1355 
1356 	return flow_hash_from_keys(keys);
1357 }
1358 EXPORT_SYMBOL(__get_hash_from_flowi6);
1359 
1360 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1361 	{
1362 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1363 		.offset = offsetof(struct flow_keys, control),
1364 	},
1365 	{
1366 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1367 		.offset = offsetof(struct flow_keys, basic),
1368 	},
1369 	{
1370 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1371 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1372 	},
1373 	{
1374 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1375 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1376 	},
1377 	{
1378 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1379 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1380 	},
1381 	{
1382 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1383 		.offset = offsetof(struct flow_keys, ports),
1384 	},
1385 	{
1386 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1387 		.offset = offsetof(struct flow_keys, vlan),
1388 	},
1389 	{
1390 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1391 		.offset = offsetof(struct flow_keys, tags),
1392 	},
1393 	{
1394 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1395 		.offset = offsetof(struct flow_keys, keyid),
1396 	},
1397 };
1398 
1399 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1400 	{
1401 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1402 		.offset = offsetof(struct flow_keys, control),
1403 	},
1404 	{
1405 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1406 		.offset = offsetof(struct flow_keys, basic),
1407 	},
1408 	{
1409 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1410 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1411 	},
1412 	{
1413 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1414 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1415 	},
1416 	{
1417 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1418 		.offset = offsetof(struct flow_keys, ports),
1419 	},
1420 };
1421 
1422 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1423 	{
1424 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1425 		.offset = offsetof(struct flow_keys, control),
1426 	},
1427 	{
1428 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1429 		.offset = offsetof(struct flow_keys, basic),
1430 	},
1431 };
1432 
1433 struct flow_dissector flow_keys_dissector __read_mostly;
1434 EXPORT_SYMBOL(flow_keys_dissector);
1435 
1436 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1437 EXPORT_SYMBOL(flow_keys_basic_dissector);
1438 
1439 static int __init init_default_flow_dissectors(void)
1440 {
1441 	skb_flow_dissector_init(&flow_keys_dissector,
1442 				flow_keys_dissector_keys,
1443 				ARRAY_SIZE(flow_keys_dissector_keys));
1444 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1445 				flow_keys_dissector_symmetric_keys,
1446 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1447 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1448 				flow_keys_basic_dissector_keys,
1449 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1450 	return 0;
1451 }
1452 
1453 core_initcall(init_default_flow_dissectors);
1454