xref: /openbmc/linux/net/core/flow_dissector.c (revision 3557b3fd)
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 #include <linux/bpf.h>
29 
30 static DEFINE_MUTEX(flow_dissector_mutex);
31 
32 static void dissector_set_key(struct flow_dissector *flow_dissector,
33 			      enum flow_dissector_key_id key_id)
34 {
35 	flow_dissector->used_keys |= (1 << key_id);
36 }
37 
38 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
39 			     const struct flow_dissector_key *key,
40 			     unsigned int key_count)
41 {
42 	unsigned int i;
43 
44 	memset(flow_dissector, 0, sizeof(*flow_dissector));
45 
46 	for (i = 0; i < key_count; i++, key++) {
47 		/* User should make sure that every key target offset is withing
48 		 * boundaries of unsigned short.
49 		 */
50 		BUG_ON(key->offset > USHRT_MAX);
51 		BUG_ON(dissector_uses_key(flow_dissector,
52 					  key->key_id));
53 
54 		dissector_set_key(flow_dissector, key->key_id);
55 		flow_dissector->offset[key->key_id] = key->offset;
56 	}
57 
58 	/* Ensure that the dissector always includes control and basic key.
59 	 * That way we are able to avoid handling lack of these in fast path.
60 	 */
61 	BUG_ON(!dissector_uses_key(flow_dissector,
62 				   FLOW_DISSECTOR_KEY_CONTROL));
63 	BUG_ON(!dissector_uses_key(flow_dissector,
64 				   FLOW_DISSECTOR_KEY_BASIC));
65 }
66 EXPORT_SYMBOL(skb_flow_dissector_init);
67 
68 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
69 				       struct bpf_prog *prog)
70 {
71 	struct bpf_prog *attached;
72 	struct net *net;
73 
74 	net = current->nsproxy->net_ns;
75 	mutex_lock(&flow_dissector_mutex);
76 	attached = rcu_dereference_protected(net->flow_dissector_prog,
77 					     lockdep_is_held(&flow_dissector_mutex));
78 	if (attached) {
79 		/* Only one BPF program can be attached at a time */
80 		mutex_unlock(&flow_dissector_mutex);
81 		return -EEXIST;
82 	}
83 	rcu_assign_pointer(net->flow_dissector_prog, prog);
84 	mutex_unlock(&flow_dissector_mutex);
85 	return 0;
86 }
87 
88 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
89 {
90 	struct bpf_prog *attached;
91 	struct net *net;
92 
93 	net = current->nsproxy->net_ns;
94 	mutex_lock(&flow_dissector_mutex);
95 	attached = rcu_dereference_protected(net->flow_dissector_prog,
96 					     lockdep_is_held(&flow_dissector_mutex));
97 	if (!attached) {
98 		mutex_unlock(&flow_dissector_mutex);
99 		return -ENOENT;
100 	}
101 	bpf_prog_put(attached);
102 	RCU_INIT_POINTER(net->flow_dissector_prog, NULL);
103 	mutex_unlock(&flow_dissector_mutex);
104 	return 0;
105 }
106 /**
107  * skb_flow_get_be16 - extract be16 entity
108  * @skb: sk_buff to extract from
109  * @poff: offset to extract at
110  * @data: raw buffer pointer to the packet
111  * @hlen: packet header length
112  *
113  * The function will try to retrieve a be32 entity at
114  * offset poff
115  */
116 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
117 				void *data, int hlen)
118 {
119 	__be16 *u, _u;
120 
121 	u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
122 	if (u)
123 		return *u;
124 
125 	return 0;
126 }
127 
128 /**
129  * __skb_flow_get_ports - extract the upper layer ports and return them
130  * @skb: sk_buff to extract the ports from
131  * @thoff: transport header offset
132  * @ip_proto: protocol for which to get port offset
133  * @data: raw buffer pointer to the packet, if NULL use skb->data
134  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
135  *
136  * The function will try to retrieve the ports at offset thoff + poff where poff
137  * is the protocol port offset returned from proto_ports_offset
138  */
139 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
140 			    void *data, int hlen)
141 {
142 	int poff = proto_ports_offset(ip_proto);
143 
144 	if (!data) {
145 		data = skb->data;
146 		hlen = skb_headlen(skb);
147 	}
148 
149 	if (poff >= 0) {
150 		__be32 *ports, _ports;
151 
152 		ports = __skb_header_pointer(skb, thoff + poff,
153 					     sizeof(_ports), data, hlen, &_ports);
154 		if (ports)
155 			return *ports;
156 	}
157 
158 	return 0;
159 }
160 EXPORT_SYMBOL(__skb_flow_get_ports);
161 
162 static void
163 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
164 				   struct flow_dissector *flow_dissector,
165 				   void *target_container)
166 {
167 	struct flow_dissector_key_control *ctrl;
168 
169 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
170 		return;
171 
172 	ctrl = skb_flow_dissector_target(flow_dissector,
173 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
174 					 target_container);
175 	ctrl->addr_type = type;
176 }
177 
178 void
179 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
180 			     struct flow_dissector *flow_dissector,
181 			     void *target_container)
182 {
183 	struct ip_tunnel_info *info;
184 	struct ip_tunnel_key *key;
185 
186 	/* A quick check to see if there might be something to do. */
187 	if (!dissector_uses_key(flow_dissector,
188 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
189 	    !dissector_uses_key(flow_dissector,
190 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
191 	    !dissector_uses_key(flow_dissector,
192 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
193 	    !dissector_uses_key(flow_dissector,
194 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
195 	    !dissector_uses_key(flow_dissector,
196 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
197 	    !dissector_uses_key(flow_dissector,
198 				FLOW_DISSECTOR_KEY_ENC_IP) &&
199 	    !dissector_uses_key(flow_dissector,
200 				FLOW_DISSECTOR_KEY_ENC_OPTS))
201 		return;
202 
203 	info = skb_tunnel_info(skb);
204 	if (!info)
205 		return;
206 
207 	key = &info->key;
208 
209 	switch (ip_tunnel_info_af(info)) {
210 	case AF_INET:
211 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
212 						   flow_dissector,
213 						   target_container);
214 		if (dissector_uses_key(flow_dissector,
215 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
216 			struct flow_dissector_key_ipv4_addrs *ipv4;
217 
218 			ipv4 = skb_flow_dissector_target(flow_dissector,
219 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
220 							 target_container);
221 			ipv4->src = key->u.ipv4.src;
222 			ipv4->dst = key->u.ipv4.dst;
223 		}
224 		break;
225 	case AF_INET6:
226 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
227 						   flow_dissector,
228 						   target_container);
229 		if (dissector_uses_key(flow_dissector,
230 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
231 			struct flow_dissector_key_ipv6_addrs *ipv6;
232 
233 			ipv6 = skb_flow_dissector_target(flow_dissector,
234 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
235 							 target_container);
236 			ipv6->src = key->u.ipv6.src;
237 			ipv6->dst = key->u.ipv6.dst;
238 		}
239 		break;
240 	}
241 
242 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
243 		struct flow_dissector_key_keyid *keyid;
244 
245 		keyid = skb_flow_dissector_target(flow_dissector,
246 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
247 						  target_container);
248 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
249 	}
250 
251 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
252 		struct flow_dissector_key_ports *tp;
253 
254 		tp = skb_flow_dissector_target(flow_dissector,
255 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
256 					       target_container);
257 		tp->src = key->tp_src;
258 		tp->dst = key->tp_dst;
259 	}
260 
261 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
262 		struct flow_dissector_key_ip *ip;
263 
264 		ip = skb_flow_dissector_target(flow_dissector,
265 					       FLOW_DISSECTOR_KEY_ENC_IP,
266 					       target_container);
267 		ip->tos = key->tos;
268 		ip->ttl = key->ttl;
269 	}
270 
271 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
272 		struct flow_dissector_key_enc_opts *enc_opt;
273 
274 		enc_opt = skb_flow_dissector_target(flow_dissector,
275 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
276 						    target_container);
277 
278 		if (info->options_len) {
279 			enc_opt->len = info->options_len;
280 			ip_tunnel_info_opts_get(enc_opt->data, info);
281 			enc_opt->dst_opt_type = info->key.tun_flags &
282 						TUNNEL_OPTIONS_PRESENT;
283 		}
284 	}
285 }
286 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
287 
288 static enum flow_dissect_ret
289 __skb_flow_dissect_mpls(const struct sk_buff *skb,
290 			struct flow_dissector *flow_dissector,
291 			void *target_container, void *data, int nhoff, int hlen)
292 {
293 	struct flow_dissector_key_keyid *key_keyid;
294 	struct mpls_label *hdr, _hdr[2];
295 	u32 entry, label;
296 
297 	if (!dissector_uses_key(flow_dissector,
298 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
299 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
300 		return FLOW_DISSECT_RET_OUT_GOOD;
301 
302 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
303 				   hlen, &_hdr);
304 	if (!hdr)
305 		return FLOW_DISSECT_RET_OUT_BAD;
306 
307 	entry = ntohl(hdr[0].entry);
308 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
309 
310 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
311 		struct flow_dissector_key_mpls *key_mpls;
312 
313 		key_mpls = skb_flow_dissector_target(flow_dissector,
314 						     FLOW_DISSECTOR_KEY_MPLS,
315 						     target_container);
316 		key_mpls->mpls_label = label;
317 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
318 					>> MPLS_LS_TTL_SHIFT;
319 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
320 					>> MPLS_LS_TC_SHIFT;
321 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
322 					>> MPLS_LS_S_SHIFT;
323 	}
324 
325 	if (label == MPLS_LABEL_ENTROPY) {
326 		key_keyid = skb_flow_dissector_target(flow_dissector,
327 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
328 						      target_container);
329 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
330 	}
331 	return FLOW_DISSECT_RET_OUT_GOOD;
332 }
333 
334 static enum flow_dissect_ret
335 __skb_flow_dissect_arp(const struct sk_buff *skb,
336 		       struct flow_dissector *flow_dissector,
337 		       void *target_container, void *data, int nhoff, int hlen)
338 {
339 	struct flow_dissector_key_arp *key_arp;
340 	struct {
341 		unsigned char ar_sha[ETH_ALEN];
342 		unsigned char ar_sip[4];
343 		unsigned char ar_tha[ETH_ALEN];
344 		unsigned char ar_tip[4];
345 	} *arp_eth, _arp_eth;
346 	const struct arphdr *arp;
347 	struct arphdr _arp;
348 
349 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
350 		return FLOW_DISSECT_RET_OUT_GOOD;
351 
352 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
353 				   hlen, &_arp);
354 	if (!arp)
355 		return FLOW_DISSECT_RET_OUT_BAD;
356 
357 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
358 	    arp->ar_pro != htons(ETH_P_IP) ||
359 	    arp->ar_hln != ETH_ALEN ||
360 	    arp->ar_pln != 4 ||
361 	    (arp->ar_op != htons(ARPOP_REPLY) &&
362 	     arp->ar_op != htons(ARPOP_REQUEST)))
363 		return FLOW_DISSECT_RET_OUT_BAD;
364 
365 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
366 				       sizeof(_arp_eth), data,
367 				       hlen, &_arp_eth);
368 	if (!arp_eth)
369 		return FLOW_DISSECT_RET_OUT_BAD;
370 
371 	key_arp = skb_flow_dissector_target(flow_dissector,
372 					    FLOW_DISSECTOR_KEY_ARP,
373 					    target_container);
374 
375 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
376 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
377 
378 	/* Only store the lower byte of the opcode;
379 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
380 	 */
381 	key_arp->op = ntohs(arp->ar_op) & 0xff;
382 
383 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
384 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
385 
386 	return FLOW_DISSECT_RET_OUT_GOOD;
387 }
388 
389 static enum flow_dissect_ret
390 __skb_flow_dissect_gre(const struct sk_buff *skb,
391 		       struct flow_dissector_key_control *key_control,
392 		       struct flow_dissector *flow_dissector,
393 		       void *target_container, void *data,
394 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
395 		       unsigned int flags)
396 {
397 	struct flow_dissector_key_keyid *key_keyid;
398 	struct gre_base_hdr *hdr, _hdr;
399 	int offset = 0;
400 	u16 gre_ver;
401 
402 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
403 				   data, *p_hlen, &_hdr);
404 	if (!hdr)
405 		return FLOW_DISSECT_RET_OUT_BAD;
406 
407 	/* Only look inside GRE without routing */
408 	if (hdr->flags & GRE_ROUTING)
409 		return FLOW_DISSECT_RET_OUT_GOOD;
410 
411 	/* Only look inside GRE for version 0 and 1 */
412 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
413 	if (gre_ver > 1)
414 		return FLOW_DISSECT_RET_OUT_GOOD;
415 
416 	*p_proto = hdr->protocol;
417 	if (gre_ver) {
418 		/* Version1 must be PPTP, and check the flags */
419 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
420 			return FLOW_DISSECT_RET_OUT_GOOD;
421 	}
422 
423 	offset += sizeof(struct gre_base_hdr);
424 
425 	if (hdr->flags & GRE_CSUM)
426 		offset += FIELD_SIZEOF(struct gre_full_hdr, csum) +
427 			  FIELD_SIZEOF(struct gre_full_hdr, reserved1);
428 
429 	if (hdr->flags & GRE_KEY) {
430 		const __be32 *keyid;
431 		__be32 _keyid;
432 
433 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
434 					     sizeof(_keyid),
435 					     data, *p_hlen, &_keyid);
436 		if (!keyid)
437 			return FLOW_DISSECT_RET_OUT_BAD;
438 
439 		if (dissector_uses_key(flow_dissector,
440 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
441 			key_keyid = skb_flow_dissector_target(flow_dissector,
442 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
443 							      target_container);
444 			if (gre_ver == 0)
445 				key_keyid->keyid = *keyid;
446 			else
447 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
448 		}
449 		offset += FIELD_SIZEOF(struct gre_full_hdr, key);
450 	}
451 
452 	if (hdr->flags & GRE_SEQ)
453 		offset += FIELD_SIZEOF(struct pptp_gre_header, seq);
454 
455 	if (gre_ver == 0) {
456 		if (*p_proto == htons(ETH_P_TEB)) {
457 			const struct ethhdr *eth;
458 			struct ethhdr _eth;
459 
460 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
461 						   sizeof(_eth),
462 						   data, *p_hlen, &_eth);
463 			if (!eth)
464 				return FLOW_DISSECT_RET_OUT_BAD;
465 			*p_proto = eth->h_proto;
466 			offset += sizeof(*eth);
467 
468 			/* Cap headers that we access via pointers at the
469 			 * end of the Ethernet header as our maximum alignment
470 			 * at that point is only 2 bytes.
471 			 */
472 			if (NET_IP_ALIGN)
473 				*p_hlen = *p_nhoff + offset;
474 		}
475 	} else { /* version 1, must be PPTP */
476 		u8 _ppp_hdr[PPP_HDRLEN];
477 		u8 *ppp_hdr;
478 
479 		if (hdr->flags & GRE_ACK)
480 			offset += FIELD_SIZEOF(struct pptp_gre_header, ack);
481 
482 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
483 					       sizeof(_ppp_hdr),
484 					       data, *p_hlen, _ppp_hdr);
485 		if (!ppp_hdr)
486 			return FLOW_DISSECT_RET_OUT_BAD;
487 
488 		switch (PPP_PROTOCOL(ppp_hdr)) {
489 		case PPP_IP:
490 			*p_proto = htons(ETH_P_IP);
491 			break;
492 		case PPP_IPV6:
493 			*p_proto = htons(ETH_P_IPV6);
494 			break;
495 		default:
496 			/* Could probably catch some more like MPLS */
497 			break;
498 		}
499 
500 		offset += PPP_HDRLEN;
501 	}
502 
503 	*p_nhoff += offset;
504 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
505 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
506 		return FLOW_DISSECT_RET_OUT_GOOD;
507 
508 	return FLOW_DISSECT_RET_PROTO_AGAIN;
509 }
510 
511 /**
512  * __skb_flow_dissect_batadv() - dissect batman-adv header
513  * @skb: sk_buff to with the batman-adv header
514  * @key_control: flow dissectors control key
515  * @data: raw buffer pointer to the packet, if NULL use skb->data
516  * @p_proto: pointer used to update the protocol to process next
517  * @p_nhoff: pointer used to update inner network header offset
518  * @hlen: packet header length
519  * @flags: any combination of FLOW_DISSECTOR_F_*
520  *
521  * ETH_P_BATMAN packets are tried to be dissected. Only
522  * &struct batadv_unicast packets are actually processed because they contain an
523  * inner ethernet header and are usually followed by actual network header. This
524  * allows the flow dissector to continue processing the packet.
525  *
526  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
527  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
528  *  otherwise FLOW_DISSECT_RET_OUT_BAD
529  */
530 static enum flow_dissect_ret
531 __skb_flow_dissect_batadv(const struct sk_buff *skb,
532 			  struct flow_dissector_key_control *key_control,
533 			  void *data, __be16 *p_proto, int *p_nhoff, int hlen,
534 			  unsigned int flags)
535 {
536 	struct {
537 		struct batadv_unicast_packet batadv_unicast;
538 		struct ethhdr eth;
539 	} *hdr, _hdr;
540 
541 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
542 				   &_hdr);
543 	if (!hdr)
544 		return FLOW_DISSECT_RET_OUT_BAD;
545 
546 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
547 		return FLOW_DISSECT_RET_OUT_BAD;
548 
549 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
550 		return FLOW_DISSECT_RET_OUT_BAD;
551 
552 	*p_proto = hdr->eth.h_proto;
553 	*p_nhoff += sizeof(*hdr);
554 
555 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
556 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
557 		return FLOW_DISSECT_RET_OUT_GOOD;
558 
559 	return FLOW_DISSECT_RET_PROTO_AGAIN;
560 }
561 
562 static void
563 __skb_flow_dissect_tcp(const struct sk_buff *skb,
564 		       struct flow_dissector *flow_dissector,
565 		       void *target_container, void *data, int thoff, int hlen)
566 {
567 	struct flow_dissector_key_tcp *key_tcp;
568 	struct tcphdr *th, _th;
569 
570 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
571 		return;
572 
573 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
574 	if (!th)
575 		return;
576 
577 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
578 		return;
579 
580 	key_tcp = skb_flow_dissector_target(flow_dissector,
581 					    FLOW_DISSECTOR_KEY_TCP,
582 					    target_container);
583 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
584 }
585 
586 static void
587 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
588 			struct flow_dissector *flow_dissector,
589 			void *target_container, void *data, const struct iphdr *iph)
590 {
591 	struct flow_dissector_key_ip *key_ip;
592 
593 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
594 		return;
595 
596 	key_ip = skb_flow_dissector_target(flow_dissector,
597 					   FLOW_DISSECTOR_KEY_IP,
598 					   target_container);
599 	key_ip->tos = iph->tos;
600 	key_ip->ttl = iph->ttl;
601 }
602 
603 static void
604 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
605 			struct flow_dissector *flow_dissector,
606 			void *target_container, void *data, const struct ipv6hdr *iph)
607 {
608 	struct flow_dissector_key_ip *key_ip;
609 
610 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
611 		return;
612 
613 	key_ip = skb_flow_dissector_target(flow_dissector,
614 					   FLOW_DISSECTOR_KEY_IP,
615 					   target_container);
616 	key_ip->tos = ipv6_get_dsfield(iph);
617 	key_ip->ttl = iph->hop_limit;
618 }
619 
620 /* Maximum number of protocol headers that can be parsed in
621  * __skb_flow_dissect
622  */
623 #define MAX_FLOW_DISSECT_HDRS	15
624 
625 static bool skb_flow_dissect_allowed(int *num_hdrs)
626 {
627 	++*num_hdrs;
628 
629 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
630 }
631 
632 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
633 				     struct flow_dissector *flow_dissector,
634 				     void *target_container)
635 {
636 	struct flow_dissector_key_control *key_control;
637 	struct flow_dissector_key_basic *key_basic;
638 	struct flow_dissector_key_addrs *key_addrs;
639 	struct flow_dissector_key_ports *key_ports;
640 
641 	key_control = skb_flow_dissector_target(flow_dissector,
642 						FLOW_DISSECTOR_KEY_CONTROL,
643 						target_container);
644 	key_control->thoff = flow_keys->thoff;
645 	if (flow_keys->is_frag)
646 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
647 	if (flow_keys->is_first_frag)
648 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
649 	if (flow_keys->is_encap)
650 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
651 
652 	key_basic = skb_flow_dissector_target(flow_dissector,
653 					      FLOW_DISSECTOR_KEY_BASIC,
654 					      target_container);
655 	key_basic->n_proto = flow_keys->n_proto;
656 	key_basic->ip_proto = flow_keys->ip_proto;
657 
658 	if (flow_keys->addr_proto == ETH_P_IP &&
659 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
660 		key_addrs = skb_flow_dissector_target(flow_dissector,
661 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
662 						      target_container);
663 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
664 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
665 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
666 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
667 		   dissector_uses_key(flow_dissector,
668 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
669 		key_addrs = skb_flow_dissector_target(flow_dissector,
670 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
671 						      target_container);
672 		memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src,
673 		       sizeof(key_addrs->v6addrs));
674 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
675 	}
676 
677 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) {
678 		key_ports = skb_flow_dissector_target(flow_dissector,
679 						      FLOW_DISSECTOR_KEY_PORTS,
680 						      target_container);
681 		key_ports->src = flow_keys->sport;
682 		key_ports->dst = flow_keys->dport;
683 	}
684 }
685 
686 bool __skb_flow_bpf_dissect(struct bpf_prog *prog,
687 			    const struct sk_buff *skb,
688 			    struct flow_dissector *flow_dissector,
689 			    struct bpf_flow_keys *flow_keys)
690 {
691 	struct bpf_skb_data_end cb_saved;
692 	struct bpf_skb_data_end *cb;
693 	u32 result;
694 
695 	/* Note that even though the const qualifier is discarded
696 	 * throughout the execution of the BPF program, all changes(the
697 	 * control block) are reverted after the BPF program returns.
698 	 * Therefore, __skb_flow_dissect does not alter the skb.
699 	 */
700 
701 	cb = (struct bpf_skb_data_end *)skb->cb;
702 
703 	/* Save Control Block */
704 	memcpy(&cb_saved, cb, sizeof(cb_saved));
705 	memset(cb, 0, sizeof(*cb));
706 
707 	/* Pass parameters to the BPF program */
708 	memset(flow_keys, 0, sizeof(*flow_keys));
709 	cb->qdisc_cb.flow_keys = flow_keys;
710 	flow_keys->n_proto = skb->protocol;
711 	flow_keys->nhoff = skb_network_offset(skb);
712 	flow_keys->thoff = flow_keys->nhoff;
713 
714 	bpf_compute_data_pointers((struct sk_buff *)skb);
715 	result = BPF_PROG_RUN(prog, skb);
716 
717 	/* Restore state */
718 	memcpy(cb, &cb_saved, sizeof(cb_saved));
719 
720 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff,
721 				   skb_network_offset(skb), skb->len);
722 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
723 				   flow_keys->nhoff, skb->len);
724 
725 	return result == BPF_OK;
726 }
727 
728 /**
729  * __skb_flow_dissect - extract the flow_keys struct and return it
730  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
731  * @flow_dissector: list of keys to dissect
732  * @target_container: target structure to put dissected values into
733  * @data: raw buffer pointer to the packet, if NULL use skb->data
734  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
735  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
736  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
737  * @flags: flags that control the dissection process, e.g.
738  *         FLOW_DISSECTOR_F_STOP_AT_L3.
739  *
740  * The function will try to retrieve individual keys into target specified
741  * by flow_dissector from either the skbuff or a raw buffer specified by the
742  * rest parameters.
743  *
744  * Caller must take care of zeroing target container memory.
745  */
746 bool __skb_flow_dissect(const struct sk_buff *skb,
747 			struct flow_dissector *flow_dissector,
748 			void *target_container,
749 			void *data, __be16 proto, int nhoff, int hlen,
750 			unsigned int flags)
751 {
752 	struct flow_dissector_key_control *key_control;
753 	struct flow_dissector_key_basic *key_basic;
754 	struct flow_dissector_key_addrs *key_addrs;
755 	struct flow_dissector_key_ports *key_ports;
756 	struct flow_dissector_key_icmp *key_icmp;
757 	struct flow_dissector_key_tags *key_tags;
758 	struct flow_dissector_key_vlan *key_vlan;
759 	enum flow_dissect_ret fdret;
760 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
761 	int num_hdrs = 0;
762 	u8 ip_proto = 0;
763 	bool ret;
764 
765 	if (!data) {
766 		data = skb->data;
767 		proto = skb_vlan_tag_present(skb) ?
768 			 skb->vlan_proto : skb->protocol;
769 		nhoff = skb_network_offset(skb);
770 		hlen = skb_headlen(skb);
771 #if IS_ENABLED(CONFIG_NET_DSA)
772 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
773 			const struct dsa_device_ops *ops;
774 			int offset;
775 
776 			ops = skb->dev->dsa_ptr->tag_ops;
777 			if (ops->flow_dissect &&
778 			    !ops->flow_dissect(skb, &proto, &offset)) {
779 				hlen -= offset;
780 				nhoff += offset;
781 			}
782 		}
783 #endif
784 	}
785 
786 	/* It is ensured by skb_flow_dissector_init() that control key will
787 	 * be always present.
788 	 */
789 	key_control = skb_flow_dissector_target(flow_dissector,
790 						FLOW_DISSECTOR_KEY_CONTROL,
791 						target_container);
792 
793 	/* It is ensured by skb_flow_dissector_init() that basic key will
794 	 * be always present.
795 	 */
796 	key_basic = skb_flow_dissector_target(flow_dissector,
797 					      FLOW_DISSECTOR_KEY_BASIC,
798 					      target_container);
799 
800 	if (skb) {
801 		struct bpf_flow_keys flow_keys;
802 		struct bpf_prog *attached = NULL;
803 
804 		rcu_read_lock();
805 
806 		if (skb->dev)
807 			attached = rcu_dereference(dev_net(skb->dev)->flow_dissector_prog);
808 		else if (skb->sk)
809 			attached = rcu_dereference(sock_net(skb->sk)->flow_dissector_prog);
810 		else
811 			WARN_ON_ONCE(1);
812 
813 		if (attached) {
814 			ret = __skb_flow_bpf_dissect(attached, skb,
815 						     flow_dissector,
816 						     &flow_keys);
817 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
818 						 target_container);
819 			rcu_read_unlock();
820 			return ret;
821 		}
822 		rcu_read_unlock();
823 	}
824 
825 	if (dissector_uses_key(flow_dissector,
826 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
827 		struct ethhdr *eth = eth_hdr(skb);
828 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
829 
830 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
831 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
832 							  target_container);
833 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
834 	}
835 
836 proto_again:
837 	fdret = FLOW_DISSECT_RET_CONTINUE;
838 
839 	switch (proto) {
840 	case htons(ETH_P_IP): {
841 		const struct iphdr *iph;
842 		struct iphdr _iph;
843 
844 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
845 		if (!iph || iph->ihl < 5) {
846 			fdret = FLOW_DISSECT_RET_OUT_BAD;
847 			break;
848 		}
849 
850 		nhoff += iph->ihl * 4;
851 
852 		ip_proto = iph->protocol;
853 
854 		if (dissector_uses_key(flow_dissector,
855 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
856 			key_addrs = skb_flow_dissector_target(flow_dissector,
857 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
858 							      target_container);
859 
860 			memcpy(&key_addrs->v4addrs, &iph->saddr,
861 			       sizeof(key_addrs->v4addrs));
862 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
863 		}
864 
865 		if (ip_is_fragment(iph)) {
866 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
867 
868 			if (iph->frag_off & htons(IP_OFFSET)) {
869 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
870 				break;
871 			} else {
872 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
873 				if (!(flags &
874 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
875 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
876 					break;
877 				}
878 			}
879 		}
880 
881 		__skb_flow_dissect_ipv4(skb, flow_dissector,
882 					target_container, data, iph);
883 
884 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
885 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
886 			break;
887 		}
888 
889 		break;
890 	}
891 	case htons(ETH_P_IPV6): {
892 		const struct ipv6hdr *iph;
893 		struct ipv6hdr _iph;
894 
895 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
896 		if (!iph) {
897 			fdret = FLOW_DISSECT_RET_OUT_BAD;
898 			break;
899 		}
900 
901 		ip_proto = iph->nexthdr;
902 		nhoff += sizeof(struct ipv6hdr);
903 
904 		if (dissector_uses_key(flow_dissector,
905 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
906 			key_addrs = skb_flow_dissector_target(flow_dissector,
907 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
908 							      target_container);
909 
910 			memcpy(&key_addrs->v6addrs, &iph->saddr,
911 			       sizeof(key_addrs->v6addrs));
912 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
913 		}
914 
915 		if ((dissector_uses_key(flow_dissector,
916 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
917 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
918 		    ip6_flowlabel(iph)) {
919 			__be32 flow_label = ip6_flowlabel(iph);
920 
921 			if (dissector_uses_key(flow_dissector,
922 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
923 				key_tags = skb_flow_dissector_target(flow_dissector,
924 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
925 								     target_container);
926 				key_tags->flow_label = ntohl(flow_label);
927 			}
928 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
929 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
930 				break;
931 			}
932 		}
933 
934 		__skb_flow_dissect_ipv6(skb, flow_dissector,
935 					target_container, data, iph);
936 
937 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
938 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
939 
940 		break;
941 	}
942 	case htons(ETH_P_8021AD):
943 	case htons(ETH_P_8021Q): {
944 		const struct vlan_hdr *vlan = NULL;
945 		struct vlan_hdr _vlan;
946 		__be16 saved_vlan_tpid = proto;
947 
948 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
949 		    skb && skb_vlan_tag_present(skb)) {
950 			proto = skb->protocol;
951 		} else {
952 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
953 						    data, hlen, &_vlan);
954 			if (!vlan) {
955 				fdret = FLOW_DISSECT_RET_OUT_BAD;
956 				break;
957 			}
958 
959 			proto = vlan->h_vlan_encapsulated_proto;
960 			nhoff += sizeof(*vlan);
961 		}
962 
963 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
964 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
965 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
966 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
967 		} else {
968 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
969 			break;
970 		}
971 
972 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
973 			key_vlan = skb_flow_dissector_target(flow_dissector,
974 							     dissector_vlan,
975 							     target_container);
976 
977 			if (!vlan) {
978 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
979 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
980 			} else {
981 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
982 					VLAN_VID_MASK;
983 				key_vlan->vlan_priority =
984 					(ntohs(vlan->h_vlan_TCI) &
985 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
986 			}
987 			key_vlan->vlan_tpid = saved_vlan_tpid;
988 		}
989 
990 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
991 		break;
992 	}
993 	case htons(ETH_P_PPP_SES): {
994 		struct {
995 			struct pppoe_hdr hdr;
996 			__be16 proto;
997 		} *hdr, _hdr;
998 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
999 		if (!hdr) {
1000 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1001 			break;
1002 		}
1003 
1004 		proto = hdr->proto;
1005 		nhoff += PPPOE_SES_HLEN;
1006 		switch (proto) {
1007 		case htons(PPP_IP):
1008 			proto = htons(ETH_P_IP);
1009 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1010 			break;
1011 		case htons(PPP_IPV6):
1012 			proto = htons(ETH_P_IPV6);
1013 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1014 			break;
1015 		default:
1016 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1017 			break;
1018 		}
1019 		break;
1020 	}
1021 	case htons(ETH_P_TIPC): {
1022 		struct tipc_basic_hdr *hdr, _hdr;
1023 
1024 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1025 					   data, hlen, &_hdr);
1026 		if (!hdr) {
1027 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1028 			break;
1029 		}
1030 
1031 		if (dissector_uses_key(flow_dissector,
1032 				       FLOW_DISSECTOR_KEY_TIPC)) {
1033 			key_addrs = skb_flow_dissector_target(flow_dissector,
1034 							      FLOW_DISSECTOR_KEY_TIPC,
1035 							      target_container);
1036 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1037 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1038 		}
1039 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1040 		break;
1041 	}
1042 
1043 	case htons(ETH_P_MPLS_UC):
1044 	case htons(ETH_P_MPLS_MC):
1045 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1046 						target_container, data,
1047 						nhoff, hlen);
1048 		break;
1049 	case htons(ETH_P_FCOE):
1050 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1051 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1052 			break;
1053 		}
1054 
1055 		nhoff += FCOE_HEADER_LEN;
1056 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1057 		break;
1058 
1059 	case htons(ETH_P_ARP):
1060 	case htons(ETH_P_RARP):
1061 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1062 					       target_container, data,
1063 					       nhoff, hlen);
1064 		break;
1065 
1066 	case htons(ETH_P_BATMAN):
1067 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1068 						  &proto, &nhoff, hlen, flags);
1069 		break;
1070 
1071 	default:
1072 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1073 		break;
1074 	}
1075 
1076 	/* Process result of proto processing */
1077 	switch (fdret) {
1078 	case FLOW_DISSECT_RET_OUT_GOOD:
1079 		goto out_good;
1080 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1081 		if (skb_flow_dissect_allowed(&num_hdrs))
1082 			goto proto_again;
1083 		goto out_good;
1084 	case FLOW_DISSECT_RET_CONTINUE:
1085 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1086 		break;
1087 	case FLOW_DISSECT_RET_OUT_BAD:
1088 	default:
1089 		goto out_bad;
1090 	}
1091 
1092 ip_proto_again:
1093 	fdret = FLOW_DISSECT_RET_CONTINUE;
1094 
1095 	switch (ip_proto) {
1096 	case IPPROTO_GRE:
1097 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1098 					       target_container, data,
1099 					       &proto, &nhoff, &hlen, flags);
1100 		break;
1101 
1102 	case NEXTHDR_HOP:
1103 	case NEXTHDR_ROUTING:
1104 	case NEXTHDR_DEST: {
1105 		u8 _opthdr[2], *opthdr;
1106 
1107 		if (proto != htons(ETH_P_IPV6))
1108 			break;
1109 
1110 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1111 					      data, hlen, &_opthdr);
1112 		if (!opthdr) {
1113 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1114 			break;
1115 		}
1116 
1117 		ip_proto = opthdr[0];
1118 		nhoff += (opthdr[1] + 1) << 3;
1119 
1120 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1121 		break;
1122 	}
1123 	case NEXTHDR_FRAGMENT: {
1124 		struct frag_hdr _fh, *fh;
1125 
1126 		if (proto != htons(ETH_P_IPV6))
1127 			break;
1128 
1129 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1130 					  data, hlen, &_fh);
1131 
1132 		if (!fh) {
1133 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1134 			break;
1135 		}
1136 
1137 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1138 
1139 		nhoff += sizeof(_fh);
1140 		ip_proto = fh->nexthdr;
1141 
1142 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1143 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1144 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1145 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1146 				break;
1147 			}
1148 		}
1149 
1150 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1151 		break;
1152 	}
1153 	case IPPROTO_IPIP:
1154 		proto = htons(ETH_P_IP);
1155 
1156 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1157 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1158 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1159 			break;
1160 		}
1161 
1162 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1163 		break;
1164 
1165 	case IPPROTO_IPV6:
1166 		proto = htons(ETH_P_IPV6);
1167 
1168 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1169 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1170 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1171 			break;
1172 		}
1173 
1174 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1175 		break;
1176 
1177 
1178 	case IPPROTO_MPLS:
1179 		proto = htons(ETH_P_MPLS_UC);
1180 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1181 		break;
1182 
1183 	case IPPROTO_TCP:
1184 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1185 				       data, nhoff, hlen);
1186 		break;
1187 
1188 	default:
1189 		break;
1190 	}
1191 
1192 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) &&
1193 	    !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) {
1194 		key_ports = skb_flow_dissector_target(flow_dissector,
1195 						      FLOW_DISSECTOR_KEY_PORTS,
1196 						      target_container);
1197 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1198 							data, hlen);
1199 	}
1200 
1201 	if (dissector_uses_key(flow_dissector,
1202 			       FLOW_DISSECTOR_KEY_ICMP)) {
1203 		key_icmp = skb_flow_dissector_target(flow_dissector,
1204 						     FLOW_DISSECTOR_KEY_ICMP,
1205 						     target_container);
1206 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1207 	}
1208 
1209 	/* Process result of IP proto processing */
1210 	switch (fdret) {
1211 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1212 		if (skb_flow_dissect_allowed(&num_hdrs))
1213 			goto proto_again;
1214 		break;
1215 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1216 		if (skb_flow_dissect_allowed(&num_hdrs))
1217 			goto ip_proto_again;
1218 		break;
1219 	case FLOW_DISSECT_RET_OUT_GOOD:
1220 	case FLOW_DISSECT_RET_CONTINUE:
1221 		break;
1222 	case FLOW_DISSECT_RET_OUT_BAD:
1223 	default:
1224 		goto out_bad;
1225 	}
1226 
1227 out_good:
1228 	ret = true;
1229 
1230 out:
1231 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1232 	key_basic->n_proto = proto;
1233 	key_basic->ip_proto = ip_proto;
1234 
1235 	return ret;
1236 
1237 out_bad:
1238 	ret = false;
1239 	goto out;
1240 }
1241 EXPORT_SYMBOL(__skb_flow_dissect);
1242 
1243 static u32 hashrnd __read_mostly;
1244 static __always_inline void __flow_hash_secret_init(void)
1245 {
1246 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1247 }
1248 
1249 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1250 					     u32 keyval)
1251 {
1252 	return jhash2(words, length, keyval);
1253 }
1254 
1255 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1256 {
1257 	const void *p = flow;
1258 
1259 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1260 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1261 }
1262 
1263 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1264 {
1265 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1266 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1267 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1268 		     sizeof(*flow) - sizeof(flow->addrs));
1269 
1270 	switch (flow->control.addr_type) {
1271 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1272 		diff -= sizeof(flow->addrs.v4addrs);
1273 		break;
1274 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1275 		diff -= sizeof(flow->addrs.v6addrs);
1276 		break;
1277 	case FLOW_DISSECTOR_KEY_TIPC:
1278 		diff -= sizeof(flow->addrs.tipckey);
1279 		break;
1280 	}
1281 	return (sizeof(*flow) - diff) / sizeof(u32);
1282 }
1283 
1284 __be32 flow_get_u32_src(const struct flow_keys *flow)
1285 {
1286 	switch (flow->control.addr_type) {
1287 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1288 		return flow->addrs.v4addrs.src;
1289 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1290 		return (__force __be32)ipv6_addr_hash(
1291 			&flow->addrs.v6addrs.src);
1292 	case FLOW_DISSECTOR_KEY_TIPC:
1293 		return flow->addrs.tipckey.key;
1294 	default:
1295 		return 0;
1296 	}
1297 }
1298 EXPORT_SYMBOL(flow_get_u32_src);
1299 
1300 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1301 {
1302 	switch (flow->control.addr_type) {
1303 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1304 		return flow->addrs.v4addrs.dst;
1305 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1306 		return (__force __be32)ipv6_addr_hash(
1307 			&flow->addrs.v6addrs.dst);
1308 	default:
1309 		return 0;
1310 	}
1311 }
1312 EXPORT_SYMBOL(flow_get_u32_dst);
1313 
1314 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1315 {
1316 	int addr_diff, i;
1317 
1318 	switch (keys->control.addr_type) {
1319 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1320 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1321 			    (__force u32)keys->addrs.v4addrs.src;
1322 		if ((addr_diff < 0) ||
1323 		    (addr_diff == 0 &&
1324 		     ((__force u16)keys->ports.dst <
1325 		      (__force u16)keys->ports.src))) {
1326 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1327 			swap(keys->ports.src, keys->ports.dst);
1328 		}
1329 		break;
1330 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1331 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1332 				   &keys->addrs.v6addrs.src,
1333 				   sizeof(keys->addrs.v6addrs.dst));
1334 		if ((addr_diff < 0) ||
1335 		    (addr_diff == 0 &&
1336 		     ((__force u16)keys->ports.dst <
1337 		      (__force u16)keys->ports.src))) {
1338 			for (i = 0; i < 4; i++)
1339 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1340 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1341 			swap(keys->ports.src, keys->ports.dst);
1342 		}
1343 		break;
1344 	}
1345 }
1346 
1347 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1348 {
1349 	u32 hash;
1350 
1351 	__flow_hash_consistentify(keys);
1352 
1353 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1354 				 flow_keys_hash_length(keys), keyval);
1355 	if (!hash)
1356 		hash = 1;
1357 
1358 	return hash;
1359 }
1360 
1361 u32 flow_hash_from_keys(struct flow_keys *keys)
1362 {
1363 	__flow_hash_secret_init();
1364 	return __flow_hash_from_keys(keys, hashrnd);
1365 }
1366 EXPORT_SYMBOL(flow_hash_from_keys);
1367 
1368 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1369 				  struct flow_keys *keys, u32 keyval)
1370 {
1371 	skb_flow_dissect_flow_keys(skb, keys,
1372 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1373 
1374 	return __flow_hash_from_keys(keys, keyval);
1375 }
1376 
1377 struct _flow_keys_digest_data {
1378 	__be16	n_proto;
1379 	u8	ip_proto;
1380 	u8	padding;
1381 	__be32	ports;
1382 	__be32	src;
1383 	__be32	dst;
1384 };
1385 
1386 void make_flow_keys_digest(struct flow_keys_digest *digest,
1387 			   const struct flow_keys *flow)
1388 {
1389 	struct _flow_keys_digest_data *data =
1390 	    (struct _flow_keys_digest_data *)digest;
1391 
1392 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1393 
1394 	memset(digest, 0, sizeof(*digest));
1395 
1396 	data->n_proto = flow->basic.n_proto;
1397 	data->ip_proto = flow->basic.ip_proto;
1398 	data->ports = flow->ports.ports;
1399 	data->src = flow->addrs.v4addrs.src;
1400 	data->dst = flow->addrs.v4addrs.dst;
1401 }
1402 EXPORT_SYMBOL(make_flow_keys_digest);
1403 
1404 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1405 
1406 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1407 {
1408 	struct flow_keys keys;
1409 
1410 	__flow_hash_secret_init();
1411 
1412 	memset(&keys, 0, sizeof(keys));
1413 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1414 			   NULL, 0, 0, 0,
1415 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1416 
1417 	return __flow_hash_from_keys(&keys, hashrnd);
1418 }
1419 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1420 
1421 /**
1422  * __skb_get_hash: calculate a flow hash
1423  * @skb: sk_buff to calculate flow hash from
1424  *
1425  * This function calculates a flow hash based on src/dst addresses
1426  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1427  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1428  * if hash is a canonical 4-tuple hash over transport ports.
1429  */
1430 void __skb_get_hash(struct sk_buff *skb)
1431 {
1432 	struct flow_keys keys;
1433 	u32 hash;
1434 
1435 	__flow_hash_secret_init();
1436 
1437 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1438 
1439 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1440 }
1441 EXPORT_SYMBOL(__skb_get_hash);
1442 
1443 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1444 {
1445 	struct flow_keys keys;
1446 
1447 	return ___skb_get_hash(skb, &keys, perturb);
1448 }
1449 EXPORT_SYMBOL(skb_get_hash_perturb);
1450 
1451 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1452 		   const struct flow_keys_basic *keys, int hlen)
1453 {
1454 	u32 poff = keys->control.thoff;
1455 
1456 	/* skip L4 headers for fragments after the first */
1457 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1458 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1459 		return poff;
1460 
1461 	switch (keys->basic.ip_proto) {
1462 	case IPPROTO_TCP: {
1463 		/* access doff as u8 to avoid unaligned access */
1464 		const u8 *doff;
1465 		u8 _doff;
1466 
1467 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1468 					    data, hlen, &_doff);
1469 		if (!doff)
1470 			return poff;
1471 
1472 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1473 		break;
1474 	}
1475 	case IPPROTO_UDP:
1476 	case IPPROTO_UDPLITE:
1477 		poff += sizeof(struct udphdr);
1478 		break;
1479 	/* For the rest, we do not really care about header
1480 	 * extensions at this point for now.
1481 	 */
1482 	case IPPROTO_ICMP:
1483 		poff += sizeof(struct icmphdr);
1484 		break;
1485 	case IPPROTO_ICMPV6:
1486 		poff += sizeof(struct icmp6hdr);
1487 		break;
1488 	case IPPROTO_IGMP:
1489 		poff += sizeof(struct igmphdr);
1490 		break;
1491 	case IPPROTO_DCCP:
1492 		poff += sizeof(struct dccp_hdr);
1493 		break;
1494 	case IPPROTO_SCTP:
1495 		poff += sizeof(struct sctphdr);
1496 		break;
1497 	}
1498 
1499 	return poff;
1500 }
1501 
1502 /**
1503  * skb_get_poff - get the offset to the payload
1504  * @skb: sk_buff to get the payload offset from
1505  *
1506  * The function will get the offset to the payload as far as it could
1507  * be dissected.  The main user is currently BPF, so that we can dynamically
1508  * truncate packets without needing to push actual payload to the user
1509  * space and can analyze headers only, instead.
1510  */
1511 u32 skb_get_poff(const struct sk_buff *skb)
1512 {
1513 	struct flow_keys_basic keys;
1514 
1515 	if (!skb_flow_dissect_flow_keys_basic(skb, &keys, NULL, 0, 0, 0, 0))
1516 		return 0;
1517 
1518 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1519 }
1520 
1521 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1522 {
1523 	memset(keys, 0, sizeof(*keys));
1524 
1525 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1526 	    sizeof(keys->addrs.v6addrs.src));
1527 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1528 	    sizeof(keys->addrs.v6addrs.dst));
1529 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1530 	keys->ports.src = fl6->fl6_sport;
1531 	keys->ports.dst = fl6->fl6_dport;
1532 	keys->keyid.keyid = fl6->fl6_gre_key;
1533 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1534 	keys->basic.ip_proto = fl6->flowi6_proto;
1535 
1536 	return flow_hash_from_keys(keys);
1537 }
1538 EXPORT_SYMBOL(__get_hash_from_flowi6);
1539 
1540 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1541 	{
1542 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1543 		.offset = offsetof(struct flow_keys, control),
1544 	},
1545 	{
1546 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1547 		.offset = offsetof(struct flow_keys, basic),
1548 	},
1549 	{
1550 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1551 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1552 	},
1553 	{
1554 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1555 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1556 	},
1557 	{
1558 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1559 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1560 	},
1561 	{
1562 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1563 		.offset = offsetof(struct flow_keys, ports),
1564 	},
1565 	{
1566 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1567 		.offset = offsetof(struct flow_keys, vlan),
1568 	},
1569 	{
1570 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1571 		.offset = offsetof(struct flow_keys, tags),
1572 	},
1573 	{
1574 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1575 		.offset = offsetof(struct flow_keys, keyid),
1576 	},
1577 };
1578 
1579 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1580 	{
1581 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1582 		.offset = offsetof(struct flow_keys, control),
1583 	},
1584 	{
1585 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1586 		.offset = offsetof(struct flow_keys, basic),
1587 	},
1588 	{
1589 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1590 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1591 	},
1592 	{
1593 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1594 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1595 	},
1596 	{
1597 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1598 		.offset = offsetof(struct flow_keys, ports),
1599 	},
1600 };
1601 
1602 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1603 	{
1604 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1605 		.offset = offsetof(struct flow_keys, control),
1606 	},
1607 	{
1608 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1609 		.offset = offsetof(struct flow_keys, basic),
1610 	},
1611 };
1612 
1613 struct flow_dissector flow_keys_dissector __read_mostly;
1614 EXPORT_SYMBOL(flow_keys_dissector);
1615 
1616 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1617 EXPORT_SYMBOL(flow_keys_basic_dissector);
1618 
1619 static int __init init_default_flow_dissectors(void)
1620 {
1621 	skb_flow_dissector_init(&flow_keys_dissector,
1622 				flow_keys_dissector_keys,
1623 				ARRAY_SIZE(flow_keys_dissector_keys));
1624 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1625 				flow_keys_dissector_symmetric_keys,
1626 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1627 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1628 				flow_keys_basic_dissector_keys,
1629 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1630 	return 0;
1631 }
1632 
1633 core_initcall(init_default_flow_dissectors);
1634