xref: /openbmc/linux/net/core/flow_dissector.c (revision a0ae2562c6c4b2721d9fddba63b7286c13517d9f)
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 		return;
157 
158 	info = skb_tunnel_info(skb);
159 	if (!info)
160 		return;
161 
162 	key = &info->key;
163 
164 	switch (ip_tunnel_info_af(info)) {
165 	case AF_INET:
166 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
167 						   flow_dissector,
168 						   target_container);
169 		if (dissector_uses_key(flow_dissector,
170 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
171 			struct flow_dissector_key_ipv4_addrs *ipv4;
172 
173 			ipv4 = skb_flow_dissector_target(flow_dissector,
174 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
175 							 target_container);
176 			ipv4->src = key->u.ipv4.src;
177 			ipv4->dst = key->u.ipv4.dst;
178 		}
179 		break;
180 	case AF_INET6:
181 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
182 						   flow_dissector,
183 						   target_container);
184 		if (dissector_uses_key(flow_dissector,
185 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
186 			struct flow_dissector_key_ipv6_addrs *ipv6;
187 
188 			ipv6 = skb_flow_dissector_target(flow_dissector,
189 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
190 							 target_container);
191 			ipv6->src = key->u.ipv6.src;
192 			ipv6->dst = key->u.ipv6.dst;
193 		}
194 		break;
195 	}
196 
197 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
198 		struct flow_dissector_key_keyid *keyid;
199 
200 		keyid = skb_flow_dissector_target(flow_dissector,
201 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
202 						  target_container);
203 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
204 	}
205 
206 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
207 		struct flow_dissector_key_ports *tp;
208 
209 		tp = skb_flow_dissector_target(flow_dissector,
210 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
211 					       target_container);
212 		tp->src = key->tp_src;
213 		tp->dst = key->tp_dst;
214 	}
215 }
216 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
217 
218 static enum flow_dissect_ret
219 __skb_flow_dissect_mpls(const struct sk_buff *skb,
220 			struct flow_dissector *flow_dissector,
221 			void *target_container, void *data, int nhoff, int hlen)
222 {
223 	struct flow_dissector_key_keyid *key_keyid;
224 	struct mpls_label *hdr, _hdr[2];
225 	u32 entry, label;
226 
227 	if (!dissector_uses_key(flow_dissector,
228 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
229 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
230 		return FLOW_DISSECT_RET_OUT_GOOD;
231 
232 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
233 				   hlen, &_hdr);
234 	if (!hdr)
235 		return FLOW_DISSECT_RET_OUT_BAD;
236 
237 	entry = ntohl(hdr[0].entry);
238 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
239 
240 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
241 		struct flow_dissector_key_mpls *key_mpls;
242 
243 		key_mpls = skb_flow_dissector_target(flow_dissector,
244 						     FLOW_DISSECTOR_KEY_MPLS,
245 						     target_container);
246 		key_mpls->mpls_label = label;
247 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
248 					>> MPLS_LS_TTL_SHIFT;
249 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
250 					>> MPLS_LS_TC_SHIFT;
251 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
252 					>> MPLS_LS_S_SHIFT;
253 	}
254 
255 	if (label == MPLS_LABEL_ENTROPY) {
256 		key_keyid = skb_flow_dissector_target(flow_dissector,
257 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
258 						      target_container);
259 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
260 	}
261 	return FLOW_DISSECT_RET_OUT_GOOD;
262 }
263 
264 static enum flow_dissect_ret
265 __skb_flow_dissect_arp(const struct sk_buff *skb,
266 		       struct flow_dissector *flow_dissector,
267 		       void *target_container, void *data, int nhoff, int hlen)
268 {
269 	struct flow_dissector_key_arp *key_arp;
270 	struct {
271 		unsigned char ar_sha[ETH_ALEN];
272 		unsigned char ar_sip[4];
273 		unsigned char ar_tha[ETH_ALEN];
274 		unsigned char ar_tip[4];
275 	} *arp_eth, _arp_eth;
276 	const struct arphdr *arp;
277 	struct arphdr _arp;
278 
279 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
280 		return FLOW_DISSECT_RET_OUT_GOOD;
281 
282 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
283 				   hlen, &_arp);
284 	if (!arp)
285 		return FLOW_DISSECT_RET_OUT_BAD;
286 
287 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
288 	    arp->ar_pro != htons(ETH_P_IP) ||
289 	    arp->ar_hln != ETH_ALEN ||
290 	    arp->ar_pln != 4 ||
291 	    (arp->ar_op != htons(ARPOP_REPLY) &&
292 	     arp->ar_op != htons(ARPOP_REQUEST)))
293 		return FLOW_DISSECT_RET_OUT_BAD;
294 
295 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
296 				       sizeof(_arp_eth), data,
297 				       hlen, &_arp_eth);
298 	if (!arp_eth)
299 		return FLOW_DISSECT_RET_OUT_BAD;
300 
301 	key_arp = skb_flow_dissector_target(flow_dissector,
302 					    FLOW_DISSECTOR_KEY_ARP,
303 					    target_container);
304 
305 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
306 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
307 
308 	/* Only store the lower byte of the opcode;
309 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
310 	 */
311 	key_arp->op = ntohs(arp->ar_op) & 0xff;
312 
313 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
314 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
315 
316 	return FLOW_DISSECT_RET_OUT_GOOD;
317 }
318 
319 static enum flow_dissect_ret
320 __skb_flow_dissect_gre(const struct sk_buff *skb,
321 		       struct flow_dissector_key_control *key_control,
322 		       struct flow_dissector *flow_dissector,
323 		       void *target_container, void *data,
324 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
325 		       unsigned int flags)
326 {
327 	struct flow_dissector_key_keyid *key_keyid;
328 	struct gre_base_hdr *hdr, _hdr;
329 	int offset = 0;
330 	u16 gre_ver;
331 
332 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
333 				   data, *p_hlen, &_hdr);
334 	if (!hdr)
335 		return FLOW_DISSECT_RET_OUT_BAD;
336 
337 	/* Only look inside GRE without routing */
338 	if (hdr->flags & GRE_ROUTING)
339 		return FLOW_DISSECT_RET_OUT_GOOD;
340 
341 	/* Only look inside GRE for version 0 and 1 */
342 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
343 	if (gre_ver > 1)
344 		return FLOW_DISSECT_RET_OUT_GOOD;
345 
346 	*p_proto = hdr->protocol;
347 	if (gre_ver) {
348 		/* Version1 must be PPTP, and check the flags */
349 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
350 			return FLOW_DISSECT_RET_OUT_GOOD;
351 	}
352 
353 	offset += sizeof(struct gre_base_hdr);
354 
355 	if (hdr->flags & GRE_CSUM)
356 		offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
357 			  sizeof(((struct gre_full_hdr *) 0)->reserved1);
358 
359 	if (hdr->flags & GRE_KEY) {
360 		const __be32 *keyid;
361 		__be32 _keyid;
362 
363 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
364 					     sizeof(_keyid),
365 					     data, *p_hlen, &_keyid);
366 		if (!keyid)
367 			return FLOW_DISSECT_RET_OUT_BAD;
368 
369 		if (dissector_uses_key(flow_dissector,
370 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
371 			key_keyid = skb_flow_dissector_target(flow_dissector,
372 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
373 							      target_container);
374 			if (gre_ver == 0)
375 				key_keyid->keyid = *keyid;
376 			else
377 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
378 		}
379 		offset += sizeof(((struct gre_full_hdr *) 0)->key);
380 	}
381 
382 	if (hdr->flags & GRE_SEQ)
383 		offset += sizeof(((struct pptp_gre_header *) 0)->seq);
384 
385 	if (gre_ver == 0) {
386 		if (*p_proto == htons(ETH_P_TEB)) {
387 			const struct ethhdr *eth;
388 			struct ethhdr _eth;
389 
390 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
391 						   sizeof(_eth),
392 						   data, *p_hlen, &_eth);
393 			if (!eth)
394 				return FLOW_DISSECT_RET_OUT_BAD;
395 			*p_proto = eth->h_proto;
396 			offset += sizeof(*eth);
397 
398 			/* Cap headers that we access via pointers at the
399 			 * end of the Ethernet header as our maximum alignment
400 			 * at that point is only 2 bytes.
401 			 */
402 			if (NET_IP_ALIGN)
403 				*p_hlen = *p_nhoff + offset;
404 		}
405 	} else { /* version 1, must be PPTP */
406 		u8 _ppp_hdr[PPP_HDRLEN];
407 		u8 *ppp_hdr;
408 
409 		if (hdr->flags & GRE_ACK)
410 			offset += sizeof(((struct pptp_gre_header *) 0)->ack);
411 
412 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
413 					       sizeof(_ppp_hdr),
414 					       data, *p_hlen, _ppp_hdr);
415 		if (!ppp_hdr)
416 			return FLOW_DISSECT_RET_OUT_BAD;
417 
418 		switch (PPP_PROTOCOL(ppp_hdr)) {
419 		case PPP_IP:
420 			*p_proto = htons(ETH_P_IP);
421 			break;
422 		case PPP_IPV6:
423 			*p_proto = htons(ETH_P_IPV6);
424 			break;
425 		default:
426 			/* Could probably catch some more like MPLS */
427 			break;
428 		}
429 
430 		offset += PPP_HDRLEN;
431 	}
432 
433 	*p_nhoff += offset;
434 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
435 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
436 		return FLOW_DISSECT_RET_OUT_GOOD;
437 
438 	return FLOW_DISSECT_RET_PROTO_AGAIN;
439 }
440 
441 /**
442  * __skb_flow_dissect_batadv() - dissect batman-adv header
443  * @skb: sk_buff to with the batman-adv header
444  * @key_control: flow dissectors control key
445  * @data: raw buffer pointer to the packet, if NULL use skb->data
446  * @p_proto: pointer used to update the protocol to process next
447  * @p_nhoff: pointer used to update inner network header offset
448  * @hlen: packet header length
449  * @flags: any combination of FLOW_DISSECTOR_F_*
450  *
451  * ETH_P_BATMAN packets are tried to be dissected. Only
452  * &struct batadv_unicast packets are actually processed because they contain an
453  * inner ethernet header and are usually followed by actual network header. This
454  * allows the flow dissector to continue processing the packet.
455  *
456  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
457  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
458  *  otherwise FLOW_DISSECT_RET_OUT_BAD
459  */
460 static enum flow_dissect_ret
461 __skb_flow_dissect_batadv(const struct sk_buff *skb,
462 			  struct flow_dissector_key_control *key_control,
463 			  void *data, __be16 *p_proto, int *p_nhoff, int hlen,
464 			  unsigned int flags)
465 {
466 	struct {
467 		struct batadv_unicast_packet batadv_unicast;
468 		struct ethhdr eth;
469 	} *hdr, _hdr;
470 
471 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
472 				   &_hdr);
473 	if (!hdr)
474 		return FLOW_DISSECT_RET_OUT_BAD;
475 
476 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
477 		return FLOW_DISSECT_RET_OUT_BAD;
478 
479 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
480 		return FLOW_DISSECT_RET_OUT_BAD;
481 
482 	*p_proto = hdr->eth.h_proto;
483 	*p_nhoff += sizeof(*hdr);
484 
485 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
486 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
487 		return FLOW_DISSECT_RET_OUT_GOOD;
488 
489 	return FLOW_DISSECT_RET_PROTO_AGAIN;
490 }
491 
492 static void
493 __skb_flow_dissect_tcp(const struct sk_buff *skb,
494 		       struct flow_dissector *flow_dissector,
495 		       void *target_container, void *data, int thoff, int hlen)
496 {
497 	struct flow_dissector_key_tcp *key_tcp;
498 	struct tcphdr *th, _th;
499 
500 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
501 		return;
502 
503 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
504 	if (!th)
505 		return;
506 
507 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
508 		return;
509 
510 	key_tcp = skb_flow_dissector_target(flow_dissector,
511 					    FLOW_DISSECTOR_KEY_TCP,
512 					    target_container);
513 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
514 }
515 
516 static void
517 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
518 			struct flow_dissector *flow_dissector,
519 			void *target_container, void *data, const struct iphdr *iph)
520 {
521 	struct flow_dissector_key_ip *key_ip;
522 
523 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
524 		return;
525 
526 	key_ip = skb_flow_dissector_target(flow_dissector,
527 					   FLOW_DISSECTOR_KEY_IP,
528 					   target_container);
529 	key_ip->tos = iph->tos;
530 	key_ip->ttl = iph->ttl;
531 }
532 
533 static void
534 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
535 			struct flow_dissector *flow_dissector,
536 			void *target_container, void *data, const struct ipv6hdr *iph)
537 {
538 	struct flow_dissector_key_ip *key_ip;
539 
540 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
541 		return;
542 
543 	key_ip = skb_flow_dissector_target(flow_dissector,
544 					   FLOW_DISSECTOR_KEY_IP,
545 					   target_container);
546 	key_ip->tos = ipv6_get_dsfield(iph);
547 	key_ip->ttl = iph->hop_limit;
548 }
549 
550 /* Maximum number of protocol headers that can be parsed in
551  * __skb_flow_dissect
552  */
553 #define MAX_FLOW_DISSECT_HDRS	15
554 
555 static bool skb_flow_dissect_allowed(int *num_hdrs)
556 {
557 	++*num_hdrs;
558 
559 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
560 }
561 
562 /**
563  * __skb_flow_dissect - extract the flow_keys struct and return it
564  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
565  * @flow_dissector: list of keys to dissect
566  * @target_container: target structure to put dissected values into
567  * @data: raw buffer pointer to the packet, if NULL use skb->data
568  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
569  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
570  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
571  *
572  * The function will try to retrieve individual keys into target specified
573  * by flow_dissector from either the skbuff or a raw buffer specified by the
574  * rest parameters.
575  *
576  * Caller must take care of zeroing target container memory.
577  */
578 bool __skb_flow_dissect(const struct sk_buff *skb,
579 			struct flow_dissector *flow_dissector,
580 			void *target_container,
581 			void *data, __be16 proto, int nhoff, int hlen,
582 			unsigned int flags)
583 {
584 	struct flow_dissector_key_control *key_control;
585 	struct flow_dissector_key_basic *key_basic;
586 	struct flow_dissector_key_addrs *key_addrs;
587 	struct flow_dissector_key_ports *key_ports;
588 	struct flow_dissector_key_icmp *key_icmp;
589 	struct flow_dissector_key_tags *key_tags;
590 	struct flow_dissector_key_vlan *key_vlan;
591 	enum flow_dissect_ret fdret;
592 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
593 	int num_hdrs = 0;
594 	u8 ip_proto = 0;
595 	bool ret;
596 
597 	if (!data) {
598 		data = skb->data;
599 		proto = skb_vlan_tag_present(skb) ?
600 			 skb->vlan_proto : skb->protocol;
601 		nhoff = skb_network_offset(skb);
602 		hlen = skb_headlen(skb);
603 #if IS_ENABLED(CONFIG_NET_DSA)
604 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
605 			const struct dsa_device_ops *ops;
606 			int offset;
607 
608 			ops = skb->dev->dsa_ptr->tag_ops;
609 			if (ops->flow_dissect &&
610 			    !ops->flow_dissect(skb, &proto, &offset)) {
611 				hlen -= offset;
612 				nhoff += offset;
613 			}
614 		}
615 #endif
616 	}
617 
618 	/* It is ensured by skb_flow_dissector_init() that control key will
619 	 * be always present.
620 	 */
621 	key_control = skb_flow_dissector_target(flow_dissector,
622 						FLOW_DISSECTOR_KEY_CONTROL,
623 						target_container);
624 
625 	/* It is ensured by skb_flow_dissector_init() that basic key will
626 	 * be always present.
627 	 */
628 	key_basic = skb_flow_dissector_target(flow_dissector,
629 					      FLOW_DISSECTOR_KEY_BASIC,
630 					      target_container);
631 
632 	if (dissector_uses_key(flow_dissector,
633 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
634 		struct ethhdr *eth = eth_hdr(skb);
635 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
636 
637 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
638 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
639 							  target_container);
640 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
641 	}
642 
643 proto_again:
644 	fdret = FLOW_DISSECT_RET_CONTINUE;
645 
646 	switch (proto) {
647 	case htons(ETH_P_IP): {
648 		const struct iphdr *iph;
649 		struct iphdr _iph;
650 
651 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
652 		if (!iph || iph->ihl < 5) {
653 			fdret = FLOW_DISSECT_RET_OUT_BAD;
654 			break;
655 		}
656 
657 		nhoff += iph->ihl * 4;
658 
659 		ip_proto = iph->protocol;
660 
661 		if (dissector_uses_key(flow_dissector,
662 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
663 			key_addrs = skb_flow_dissector_target(flow_dissector,
664 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
665 							      target_container);
666 
667 			memcpy(&key_addrs->v4addrs, &iph->saddr,
668 			       sizeof(key_addrs->v4addrs));
669 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
670 		}
671 
672 		if (ip_is_fragment(iph)) {
673 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
674 
675 			if (iph->frag_off & htons(IP_OFFSET)) {
676 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
677 				break;
678 			} else {
679 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
680 				if (!(flags &
681 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
682 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
683 					break;
684 				}
685 			}
686 		}
687 
688 		__skb_flow_dissect_ipv4(skb, flow_dissector,
689 					target_container, data, iph);
690 
691 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
692 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
693 			break;
694 		}
695 
696 		break;
697 	}
698 	case htons(ETH_P_IPV6): {
699 		const struct ipv6hdr *iph;
700 		struct ipv6hdr _iph;
701 
702 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
703 		if (!iph) {
704 			fdret = FLOW_DISSECT_RET_OUT_BAD;
705 			break;
706 		}
707 
708 		ip_proto = iph->nexthdr;
709 		nhoff += sizeof(struct ipv6hdr);
710 
711 		if (dissector_uses_key(flow_dissector,
712 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
713 			key_addrs = skb_flow_dissector_target(flow_dissector,
714 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
715 							      target_container);
716 
717 			memcpy(&key_addrs->v6addrs, &iph->saddr,
718 			       sizeof(key_addrs->v6addrs));
719 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
720 		}
721 
722 		if ((dissector_uses_key(flow_dissector,
723 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
724 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
725 		    ip6_flowlabel(iph)) {
726 			__be32 flow_label = ip6_flowlabel(iph);
727 
728 			if (dissector_uses_key(flow_dissector,
729 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
730 				key_tags = skb_flow_dissector_target(flow_dissector,
731 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
732 								     target_container);
733 				key_tags->flow_label = ntohl(flow_label);
734 			}
735 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
736 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
737 				break;
738 			}
739 		}
740 
741 		__skb_flow_dissect_ipv6(skb, flow_dissector,
742 					target_container, data, iph);
743 
744 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
745 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
746 
747 		break;
748 	}
749 	case htons(ETH_P_8021AD):
750 	case htons(ETH_P_8021Q): {
751 		const struct vlan_hdr *vlan = NULL;
752 		struct vlan_hdr _vlan;
753 		__be16 saved_vlan_tpid = proto;
754 
755 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
756 		    skb && skb_vlan_tag_present(skb)) {
757 			proto = skb->protocol;
758 		} else {
759 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
760 						    data, hlen, &_vlan);
761 			if (!vlan) {
762 				fdret = FLOW_DISSECT_RET_OUT_BAD;
763 				break;
764 			}
765 
766 			proto = vlan->h_vlan_encapsulated_proto;
767 			nhoff += sizeof(*vlan);
768 		}
769 
770 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
771 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
772 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
773 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
774 		} else {
775 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
776 			break;
777 		}
778 
779 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
780 			key_vlan = skb_flow_dissector_target(flow_dissector,
781 							     dissector_vlan,
782 							     target_container);
783 
784 			if (!vlan) {
785 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
786 				key_vlan->vlan_priority =
787 					(skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
788 			} else {
789 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
790 					VLAN_VID_MASK;
791 				key_vlan->vlan_priority =
792 					(ntohs(vlan->h_vlan_TCI) &
793 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
794 			}
795 			key_vlan->vlan_tpid = saved_vlan_tpid;
796 		}
797 
798 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
799 		break;
800 	}
801 	case htons(ETH_P_PPP_SES): {
802 		struct {
803 			struct pppoe_hdr hdr;
804 			__be16 proto;
805 		} *hdr, _hdr;
806 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
807 		if (!hdr) {
808 			fdret = FLOW_DISSECT_RET_OUT_BAD;
809 			break;
810 		}
811 
812 		proto = hdr->proto;
813 		nhoff += PPPOE_SES_HLEN;
814 		switch (proto) {
815 		case htons(PPP_IP):
816 			proto = htons(ETH_P_IP);
817 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
818 			break;
819 		case htons(PPP_IPV6):
820 			proto = htons(ETH_P_IPV6);
821 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
822 			break;
823 		default:
824 			fdret = FLOW_DISSECT_RET_OUT_BAD;
825 			break;
826 		}
827 		break;
828 	}
829 	case htons(ETH_P_TIPC): {
830 		struct tipc_basic_hdr *hdr, _hdr;
831 
832 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
833 					   data, hlen, &_hdr);
834 		if (!hdr) {
835 			fdret = FLOW_DISSECT_RET_OUT_BAD;
836 			break;
837 		}
838 
839 		if (dissector_uses_key(flow_dissector,
840 				       FLOW_DISSECTOR_KEY_TIPC)) {
841 			key_addrs = skb_flow_dissector_target(flow_dissector,
842 							      FLOW_DISSECTOR_KEY_TIPC,
843 							      target_container);
844 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
845 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
846 		}
847 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
848 		break;
849 	}
850 
851 	case htons(ETH_P_MPLS_UC):
852 	case htons(ETH_P_MPLS_MC):
853 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
854 						target_container, data,
855 						nhoff, hlen);
856 		break;
857 	case htons(ETH_P_FCOE):
858 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
859 			fdret = FLOW_DISSECT_RET_OUT_BAD;
860 			break;
861 		}
862 
863 		nhoff += FCOE_HEADER_LEN;
864 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
865 		break;
866 
867 	case htons(ETH_P_ARP):
868 	case htons(ETH_P_RARP):
869 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
870 					       target_container, data,
871 					       nhoff, hlen);
872 		break;
873 
874 	case htons(ETH_P_BATMAN):
875 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
876 						  &proto, &nhoff, hlen, flags);
877 		break;
878 
879 	default:
880 		fdret = FLOW_DISSECT_RET_OUT_BAD;
881 		break;
882 	}
883 
884 	/* Process result of proto processing */
885 	switch (fdret) {
886 	case FLOW_DISSECT_RET_OUT_GOOD:
887 		goto out_good;
888 	case FLOW_DISSECT_RET_PROTO_AGAIN:
889 		if (skb_flow_dissect_allowed(&num_hdrs))
890 			goto proto_again;
891 		goto out_good;
892 	case FLOW_DISSECT_RET_CONTINUE:
893 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
894 		break;
895 	case FLOW_DISSECT_RET_OUT_BAD:
896 	default:
897 		goto out_bad;
898 	}
899 
900 ip_proto_again:
901 	fdret = FLOW_DISSECT_RET_CONTINUE;
902 
903 	switch (ip_proto) {
904 	case IPPROTO_GRE:
905 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
906 					       target_container, data,
907 					       &proto, &nhoff, &hlen, flags);
908 		break;
909 
910 	case NEXTHDR_HOP:
911 	case NEXTHDR_ROUTING:
912 	case NEXTHDR_DEST: {
913 		u8 _opthdr[2], *opthdr;
914 
915 		if (proto != htons(ETH_P_IPV6))
916 			break;
917 
918 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
919 					      data, hlen, &_opthdr);
920 		if (!opthdr) {
921 			fdret = FLOW_DISSECT_RET_OUT_BAD;
922 			break;
923 		}
924 
925 		ip_proto = opthdr[0];
926 		nhoff += (opthdr[1] + 1) << 3;
927 
928 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
929 		break;
930 	}
931 	case NEXTHDR_FRAGMENT: {
932 		struct frag_hdr _fh, *fh;
933 
934 		if (proto != htons(ETH_P_IPV6))
935 			break;
936 
937 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
938 					  data, hlen, &_fh);
939 
940 		if (!fh) {
941 			fdret = FLOW_DISSECT_RET_OUT_BAD;
942 			break;
943 		}
944 
945 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
946 
947 		nhoff += sizeof(_fh);
948 		ip_proto = fh->nexthdr;
949 
950 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
951 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
952 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
953 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
954 				break;
955 			}
956 		}
957 
958 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
959 		break;
960 	}
961 	case IPPROTO_IPIP:
962 		proto = htons(ETH_P_IP);
963 
964 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
965 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
966 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
967 			break;
968 		}
969 
970 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
971 		break;
972 
973 	case IPPROTO_IPV6:
974 		proto = htons(ETH_P_IPV6);
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 
986 	case IPPROTO_MPLS:
987 		proto = htons(ETH_P_MPLS_UC);
988 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
989 		break;
990 
991 	case IPPROTO_TCP:
992 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
993 				       data, nhoff, hlen);
994 		break;
995 
996 	default:
997 		break;
998 	}
999 
1000 	if (dissector_uses_key(flow_dissector,
1001 			       FLOW_DISSECTOR_KEY_PORTS)) {
1002 		key_ports = skb_flow_dissector_target(flow_dissector,
1003 						      FLOW_DISSECTOR_KEY_PORTS,
1004 						      target_container);
1005 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1006 							data, hlen);
1007 	}
1008 
1009 	if (dissector_uses_key(flow_dissector,
1010 			       FLOW_DISSECTOR_KEY_ICMP)) {
1011 		key_icmp = skb_flow_dissector_target(flow_dissector,
1012 						     FLOW_DISSECTOR_KEY_ICMP,
1013 						     target_container);
1014 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1015 	}
1016 
1017 	/* Process result of IP proto processing */
1018 	switch (fdret) {
1019 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1020 		if (skb_flow_dissect_allowed(&num_hdrs))
1021 			goto proto_again;
1022 		break;
1023 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1024 		if (skb_flow_dissect_allowed(&num_hdrs))
1025 			goto ip_proto_again;
1026 		break;
1027 	case FLOW_DISSECT_RET_OUT_GOOD:
1028 	case FLOW_DISSECT_RET_CONTINUE:
1029 		break;
1030 	case FLOW_DISSECT_RET_OUT_BAD:
1031 	default:
1032 		goto out_bad;
1033 	}
1034 
1035 out_good:
1036 	ret = true;
1037 
1038 out:
1039 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1040 	key_basic->n_proto = proto;
1041 	key_basic->ip_proto = ip_proto;
1042 
1043 	return ret;
1044 
1045 out_bad:
1046 	ret = false;
1047 	goto out;
1048 }
1049 EXPORT_SYMBOL(__skb_flow_dissect);
1050 
1051 static u32 hashrnd __read_mostly;
1052 static __always_inline void __flow_hash_secret_init(void)
1053 {
1054 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1055 }
1056 
1057 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1058 					     u32 keyval)
1059 {
1060 	return jhash2(words, length, keyval);
1061 }
1062 
1063 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1064 {
1065 	const void *p = flow;
1066 
1067 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1068 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1069 }
1070 
1071 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1072 {
1073 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1074 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1075 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1076 		     sizeof(*flow) - sizeof(flow->addrs));
1077 
1078 	switch (flow->control.addr_type) {
1079 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1080 		diff -= sizeof(flow->addrs.v4addrs);
1081 		break;
1082 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1083 		diff -= sizeof(flow->addrs.v6addrs);
1084 		break;
1085 	case FLOW_DISSECTOR_KEY_TIPC:
1086 		diff -= sizeof(flow->addrs.tipckey);
1087 		break;
1088 	}
1089 	return (sizeof(*flow) - diff) / sizeof(u32);
1090 }
1091 
1092 __be32 flow_get_u32_src(const struct flow_keys *flow)
1093 {
1094 	switch (flow->control.addr_type) {
1095 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1096 		return flow->addrs.v4addrs.src;
1097 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1098 		return (__force __be32)ipv6_addr_hash(
1099 			&flow->addrs.v6addrs.src);
1100 	case FLOW_DISSECTOR_KEY_TIPC:
1101 		return flow->addrs.tipckey.key;
1102 	default:
1103 		return 0;
1104 	}
1105 }
1106 EXPORT_SYMBOL(flow_get_u32_src);
1107 
1108 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1109 {
1110 	switch (flow->control.addr_type) {
1111 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1112 		return flow->addrs.v4addrs.dst;
1113 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1114 		return (__force __be32)ipv6_addr_hash(
1115 			&flow->addrs.v6addrs.dst);
1116 	default:
1117 		return 0;
1118 	}
1119 }
1120 EXPORT_SYMBOL(flow_get_u32_dst);
1121 
1122 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1123 {
1124 	int addr_diff, i;
1125 
1126 	switch (keys->control.addr_type) {
1127 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1128 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1129 			    (__force u32)keys->addrs.v4addrs.src;
1130 		if ((addr_diff < 0) ||
1131 		    (addr_diff == 0 &&
1132 		     ((__force u16)keys->ports.dst <
1133 		      (__force u16)keys->ports.src))) {
1134 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1135 			swap(keys->ports.src, keys->ports.dst);
1136 		}
1137 		break;
1138 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1139 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1140 				   &keys->addrs.v6addrs.src,
1141 				   sizeof(keys->addrs.v6addrs.dst));
1142 		if ((addr_diff < 0) ||
1143 		    (addr_diff == 0 &&
1144 		     ((__force u16)keys->ports.dst <
1145 		      (__force u16)keys->ports.src))) {
1146 			for (i = 0; i < 4; i++)
1147 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1148 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1149 			swap(keys->ports.src, keys->ports.dst);
1150 		}
1151 		break;
1152 	}
1153 }
1154 
1155 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1156 {
1157 	u32 hash;
1158 
1159 	__flow_hash_consistentify(keys);
1160 
1161 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1162 				 flow_keys_hash_length(keys), keyval);
1163 	if (!hash)
1164 		hash = 1;
1165 
1166 	return hash;
1167 }
1168 
1169 u32 flow_hash_from_keys(struct flow_keys *keys)
1170 {
1171 	__flow_hash_secret_init();
1172 	return __flow_hash_from_keys(keys, hashrnd);
1173 }
1174 EXPORT_SYMBOL(flow_hash_from_keys);
1175 
1176 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1177 				  struct flow_keys *keys, u32 keyval)
1178 {
1179 	skb_flow_dissect_flow_keys(skb, keys,
1180 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1181 
1182 	return __flow_hash_from_keys(keys, keyval);
1183 }
1184 
1185 struct _flow_keys_digest_data {
1186 	__be16	n_proto;
1187 	u8	ip_proto;
1188 	u8	padding;
1189 	__be32	ports;
1190 	__be32	src;
1191 	__be32	dst;
1192 };
1193 
1194 void make_flow_keys_digest(struct flow_keys_digest *digest,
1195 			   const struct flow_keys *flow)
1196 {
1197 	struct _flow_keys_digest_data *data =
1198 	    (struct _flow_keys_digest_data *)digest;
1199 
1200 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1201 
1202 	memset(digest, 0, sizeof(*digest));
1203 
1204 	data->n_proto = flow->basic.n_proto;
1205 	data->ip_proto = flow->basic.ip_proto;
1206 	data->ports = flow->ports.ports;
1207 	data->src = flow->addrs.v4addrs.src;
1208 	data->dst = flow->addrs.v4addrs.dst;
1209 }
1210 EXPORT_SYMBOL(make_flow_keys_digest);
1211 
1212 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1213 
1214 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1215 {
1216 	struct flow_keys keys;
1217 
1218 	__flow_hash_secret_init();
1219 
1220 	memset(&keys, 0, sizeof(keys));
1221 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1222 			   NULL, 0, 0, 0,
1223 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1224 
1225 	return __flow_hash_from_keys(&keys, hashrnd);
1226 }
1227 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1228 
1229 /**
1230  * __skb_get_hash: calculate a flow hash
1231  * @skb: sk_buff to calculate flow hash from
1232  *
1233  * This function calculates a flow hash based on src/dst addresses
1234  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1235  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1236  * if hash is a canonical 4-tuple hash over transport ports.
1237  */
1238 void __skb_get_hash(struct sk_buff *skb)
1239 {
1240 	struct flow_keys keys;
1241 	u32 hash;
1242 
1243 	__flow_hash_secret_init();
1244 
1245 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1246 
1247 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1248 }
1249 EXPORT_SYMBOL(__skb_get_hash);
1250 
1251 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1252 {
1253 	struct flow_keys keys;
1254 
1255 	return ___skb_get_hash(skb, &keys, perturb);
1256 }
1257 EXPORT_SYMBOL(skb_get_hash_perturb);
1258 
1259 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1260 		   const struct flow_keys_basic *keys, int hlen)
1261 {
1262 	u32 poff = keys->control.thoff;
1263 
1264 	/* skip L4 headers for fragments after the first */
1265 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1266 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1267 		return poff;
1268 
1269 	switch (keys->basic.ip_proto) {
1270 	case IPPROTO_TCP: {
1271 		/* access doff as u8 to avoid unaligned access */
1272 		const u8 *doff;
1273 		u8 _doff;
1274 
1275 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1276 					    data, hlen, &_doff);
1277 		if (!doff)
1278 			return poff;
1279 
1280 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1281 		break;
1282 	}
1283 	case IPPROTO_UDP:
1284 	case IPPROTO_UDPLITE:
1285 		poff += sizeof(struct udphdr);
1286 		break;
1287 	/* For the rest, we do not really care about header
1288 	 * extensions at this point for now.
1289 	 */
1290 	case IPPROTO_ICMP:
1291 		poff += sizeof(struct icmphdr);
1292 		break;
1293 	case IPPROTO_ICMPV6:
1294 		poff += sizeof(struct icmp6hdr);
1295 		break;
1296 	case IPPROTO_IGMP:
1297 		poff += sizeof(struct igmphdr);
1298 		break;
1299 	case IPPROTO_DCCP:
1300 		poff += sizeof(struct dccp_hdr);
1301 		break;
1302 	case IPPROTO_SCTP:
1303 		poff += sizeof(struct sctphdr);
1304 		break;
1305 	}
1306 
1307 	return poff;
1308 }
1309 
1310 /**
1311  * skb_get_poff - get the offset to the payload
1312  * @skb: sk_buff to get the payload offset from
1313  *
1314  * The function will get the offset to the payload as far as it could
1315  * be dissected.  The main user is currently BPF, so that we can dynamically
1316  * truncate packets without needing to push actual payload to the user
1317  * space and can analyze headers only, instead.
1318  */
1319 u32 skb_get_poff(const struct sk_buff *skb)
1320 {
1321 	struct flow_keys_basic keys;
1322 
1323 	if (!skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0))
1324 		return 0;
1325 
1326 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1327 }
1328 
1329 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1330 {
1331 	memset(keys, 0, sizeof(*keys));
1332 
1333 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1334 	    sizeof(keys->addrs.v6addrs.src));
1335 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1336 	    sizeof(keys->addrs.v6addrs.dst));
1337 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1338 	keys->ports.src = fl6->fl6_sport;
1339 	keys->ports.dst = fl6->fl6_dport;
1340 	keys->keyid.keyid = fl6->fl6_gre_key;
1341 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1342 	keys->basic.ip_proto = fl6->flowi6_proto;
1343 
1344 	return flow_hash_from_keys(keys);
1345 }
1346 EXPORT_SYMBOL(__get_hash_from_flowi6);
1347 
1348 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1349 	{
1350 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1351 		.offset = offsetof(struct flow_keys, control),
1352 	},
1353 	{
1354 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1355 		.offset = offsetof(struct flow_keys, basic),
1356 	},
1357 	{
1358 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1359 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1360 	},
1361 	{
1362 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1363 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1364 	},
1365 	{
1366 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1367 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1368 	},
1369 	{
1370 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1371 		.offset = offsetof(struct flow_keys, ports),
1372 	},
1373 	{
1374 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1375 		.offset = offsetof(struct flow_keys, vlan),
1376 	},
1377 	{
1378 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1379 		.offset = offsetof(struct flow_keys, tags),
1380 	},
1381 	{
1382 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1383 		.offset = offsetof(struct flow_keys, keyid),
1384 	},
1385 };
1386 
1387 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1388 	{
1389 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1390 		.offset = offsetof(struct flow_keys, control),
1391 	},
1392 	{
1393 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1394 		.offset = offsetof(struct flow_keys, basic),
1395 	},
1396 	{
1397 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1398 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1399 	},
1400 	{
1401 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1402 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1403 	},
1404 	{
1405 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1406 		.offset = offsetof(struct flow_keys, ports),
1407 	},
1408 };
1409 
1410 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1411 	{
1412 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1413 		.offset = offsetof(struct flow_keys, control),
1414 	},
1415 	{
1416 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1417 		.offset = offsetof(struct flow_keys, basic),
1418 	},
1419 };
1420 
1421 struct flow_dissector flow_keys_dissector __read_mostly;
1422 EXPORT_SYMBOL(flow_keys_dissector);
1423 
1424 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1425 EXPORT_SYMBOL(flow_keys_basic_dissector);
1426 
1427 static int __init init_default_flow_dissectors(void)
1428 {
1429 	skb_flow_dissector_init(&flow_keys_dissector,
1430 				flow_keys_dissector_keys,
1431 				ARRAY_SIZE(flow_keys_dissector_keys));
1432 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1433 				flow_keys_dissector_symmetric_keys,
1434 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1435 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1436 				flow_keys_basic_dissector_keys,
1437 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1438 	return 0;
1439 }
1440 
1441 core_initcall(init_default_flow_dissectors);
1442