xref: /openbmc/linux/net/core/flow_dissector.c (revision b593bce5)
1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/kernel.h>
3 #include <linux/skbuff.h>
4 #include <linux/export.h>
5 #include <linux/ip.h>
6 #include <linux/ipv6.h>
7 #include <linux/if_vlan.h>
8 #include <net/dsa.h>
9 #include <net/dst_metadata.h>
10 #include <net/ip.h>
11 #include <net/ipv6.h>
12 #include <net/gre.h>
13 #include <net/pptp.h>
14 #include <net/tipc.h>
15 #include <linux/igmp.h>
16 #include <linux/icmp.h>
17 #include <linux/sctp.h>
18 #include <linux/dccp.h>
19 #include <linux/if_tunnel.h>
20 #include <linux/if_pppox.h>
21 #include <linux/ppp_defs.h>
22 #include <linux/stddef.h>
23 #include <linux/if_ether.h>
24 #include <linux/mpls.h>
25 #include <linux/tcp.h>
26 #include <net/flow_dissector.h>
27 #include <scsi/fc/fc_fcoe.h>
28 #include <uapi/linux/batadv_packet.h>
29 #include <linux/bpf.h>
30 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
31 #include <net/netfilter/nf_conntrack_core.h>
32 #include <net/netfilter/nf_conntrack_labels.h>
33 #endif
34 
35 static DEFINE_MUTEX(flow_dissector_mutex);
36 
37 static void dissector_set_key(struct flow_dissector *flow_dissector,
38 			      enum flow_dissector_key_id key_id)
39 {
40 	flow_dissector->used_keys |= (1 << key_id);
41 }
42 
43 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
44 			     const struct flow_dissector_key *key,
45 			     unsigned int key_count)
46 {
47 	unsigned int i;
48 
49 	memset(flow_dissector, 0, sizeof(*flow_dissector));
50 
51 	for (i = 0; i < key_count; i++, key++) {
52 		/* User should make sure that every key target offset is withing
53 		 * boundaries of unsigned short.
54 		 */
55 		BUG_ON(key->offset > USHRT_MAX);
56 		BUG_ON(dissector_uses_key(flow_dissector,
57 					  key->key_id));
58 
59 		dissector_set_key(flow_dissector, key->key_id);
60 		flow_dissector->offset[key->key_id] = key->offset;
61 	}
62 
63 	/* Ensure that the dissector always includes control and basic key.
64 	 * That way we are able to avoid handling lack of these in fast path.
65 	 */
66 	BUG_ON(!dissector_uses_key(flow_dissector,
67 				   FLOW_DISSECTOR_KEY_CONTROL));
68 	BUG_ON(!dissector_uses_key(flow_dissector,
69 				   FLOW_DISSECTOR_KEY_BASIC));
70 }
71 EXPORT_SYMBOL(skb_flow_dissector_init);
72 
73 int skb_flow_dissector_prog_query(const union bpf_attr *attr,
74 				  union bpf_attr __user *uattr)
75 {
76 	__u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids);
77 	u32 prog_id, prog_cnt = 0, flags = 0;
78 	struct bpf_prog *attached;
79 	struct net *net;
80 
81 	if (attr->query.query_flags)
82 		return -EINVAL;
83 
84 	net = get_net_ns_by_fd(attr->query.target_fd);
85 	if (IS_ERR(net))
86 		return PTR_ERR(net);
87 
88 	rcu_read_lock();
89 	attached = rcu_dereference(net->flow_dissector_prog);
90 	if (attached) {
91 		prog_cnt = 1;
92 		prog_id = attached->aux->id;
93 	}
94 	rcu_read_unlock();
95 
96 	put_net(net);
97 
98 	if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
99 		return -EFAULT;
100 	if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt)))
101 		return -EFAULT;
102 
103 	if (!attr->query.prog_cnt || !prog_ids || !prog_cnt)
104 		return 0;
105 
106 	if (copy_to_user(prog_ids, &prog_id, sizeof(u32)))
107 		return -EFAULT;
108 
109 	return 0;
110 }
111 
112 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
113 				       struct bpf_prog *prog)
114 {
115 	struct bpf_prog *attached;
116 	struct net *net;
117 
118 	net = current->nsproxy->net_ns;
119 	mutex_lock(&flow_dissector_mutex);
120 	attached = rcu_dereference_protected(net->flow_dissector_prog,
121 					     lockdep_is_held(&flow_dissector_mutex));
122 	if (attached) {
123 		/* Only one BPF program can be attached at a time */
124 		mutex_unlock(&flow_dissector_mutex);
125 		return -EEXIST;
126 	}
127 	rcu_assign_pointer(net->flow_dissector_prog, prog);
128 	mutex_unlock(&flow_dissector_mutex);
129 	return 0;
130 }
131 
132 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
133 {
134 	struct bpf_prog *attached;
135 	struct net *net;
136 
137 	net = current->nsproxy->net_ns;
138 	mutex_lock(&flow_dissector_mutex);
139 	attached = rcu_dereference_protected(net->flow_dissector_prog,
140 					     lockdep_is_held(&flow_dissector_mutex));
141 	if (!attached) {
142 		mutex_unlock(&flow_dissector_mutex);
143 		return -ENOENT;
144 	}
145 	bpf_prog_put(attached);
146 	RCU_INIT_POINTER(net->flow_dissector_prog, NULL);
147 	mutex_unlock(&flow_dissector_mutex);
148 	return 0;
149 }
150 /**
151  * skb_flow_get_be16 - extract be16 entity
152  * @skb: sk_buff to extract from
153  * @poff: offset to extract at
154  * @data: raw buffer pointer to the packet
155  * @hlen: packet header length
156  *
157  * The function will try to retrieve a be32 entity at
158  * offset poff
159  */
160 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
161 				void *data, int hlen)
162 {
163 	__be16 *u, _u;
164 
165 	u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
166 	if (u)
167 		return *u;
168 
169 	return 0;
170 }
171 
172 /**
173  * __skb_flow_get_ports - extract the upper layer ports and return them
174  * @skb: sk_buff to extract the ports from
175  * @thoff: transport header offset
176  * @ip_proto: protocol for which to get port offset
177  * @data: raw buffer pointer to the packet, if NULL use skb->data
178  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
179  *
180  * The function will try to retrieve the ports at offset thoff + poff where poff
181  * is the protocol port offset returned from proto_ports_offset
182  */
183 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
184 			    void *data, int hlen)
185 {
186 	int poff = proto_ports_offset(ip_proto);
187 
188 	if (!data) {
189 		data = skb->data;
190 		hlen = skb_headlen(skb);
191 	}
192 
193 	if (poff >= 0) {
194 		__be32 *ports, _ports;
195 
196 		ports = __skb_header_pointer(skb, thoff + poff,
197 					     sizeof(_ports), data, hlen, &_ports);
198 		if (ports)
199 			return *ports;
200 	}
201 
202 	return 0;
203 }
204 EXPORT_SYMBOL(__skb_flow_get_ports);
205 
206 void skb_flow_dissect_meta(const struct sk_buff *skb,
207 			   struct flow_dissector *flow_dissector,
208 			   void *target_container)
209 {
210 	struct flow_dissector_key_meta *meta;
211 
212 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
213 		return;
214 
215 	meta = skb_flow_dissector_target(flow_dissector,
216 					 FLOW_DISSECTOR_KEY_META,
217 					 target_container);
218 	meta->ingress_ifindex = skb->skb_iif;
219 }
220 EXPORT_SYMBOL(skb_flow_dissect_meta);
221 
222 static void
223 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
224 				   struct flow_dissector *flow_dissector,
225 				   void *target_container)
226 {
227 	struct flow_dissector_key_control *ctrl;
228 
229 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
230 		return;
231 
232 	ctrl = skb_flow_dissector_target(flow_dissector,
233 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
234 					 target_container);
235 	ctrl->addr_type = type;
236 }
237 
238 void
239 skb_flow_dissect_ct(const struct sk_buff *skb,
240 		    struct flow_dissector *flow_dissector,
241 		    void *target_container,
242 		    u16 *ctinfo_map,
243 		    size_t mapsize)
244 {
245 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
246 	struct flow_dissector_key_ct *key;
247 	enum ip_conntrack_info ctinfo;
248 	struct nf_conn_labels *cl;
249 	struct nf_conn *ct;
250 
251 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
252 		return;
253 
254 	ct = nf_ct_get(skb, &ctinfo);
255 	if (!ct)
256 		return;
257 
258 	key = skb_flow_dissector_target(flow_dissector,
259 					FLOW_DISSECTOR_KEY_CT,
260 					target_container);
261 
262 	if (ctinfo < mapsize)
263 		key->ct_state = ctinfo_map[ctinfo];
264 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
265 	key->ct_zone = ct->zone.id;
266 #endif
267 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
268 	key->ct_mark = ct->mark;
269 #endif
270 
271 	cl = nf_ct_labels_find(ct);
272 	if (cl)
273 		memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
274 #endif /* CONFIG_NF_CONNTRACK */
275 }
276 EXPORT_SYMBOL(skb_flow_dissect_ct);
277 
278 void
279 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
280 			     struct flow_dissector *flow_dissector,
281 			     void *target_container)
282 {
283 	struct ip_tunnel_info *info;
284 	struct ip_tunnel_key *key;
285 
286 	/* A quick check to see if there might be something to do. */
287 	if (!dissector_uses_key(flow_dissector,
288 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
289 	    !dissector_uses_key(flow_dissector,
290 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
291 	    !dissector_uses_key(flow_dissector,
292 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
293 	    !dissector_uses_key(flow_dissector,
294 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
295 	    !dissector_uses_key(flow_dissector,
296 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
297 	    !dissector_uses_key(flow_dissector,
298 				FLOW_DISSECTOR_KEY_ENC_IP) &&
299 	    !dissector_uses_key(flow_dissector,
300 				FLOW_DISSECTOR_KEY_ENC_OPTS))
301 		return;
302 
303 	info = skb_tunnel_info(skb);
304 	if (!info)
305 		return;
306 
307 	key = &info->key;
308 
309 	switch (ip_tunnel_info_af(info)) {
310 	case AF_INET:
311 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
312 						   flow_dissector,
313 						   target_container);
314 		if (dissector_uses_key(flow_dissector,
315 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
316 			struct flow_dissector_key_ipv4_addrs *ipv4;
317 
318 			ipv4 = skb_flow_dissector_target(flow_dissector,
319 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
320 							 target_container);
321 			ipv4->src = key->u.ipv4.src;
322 			ipv4->dst = key->u.ipv4.dst;
323 		}
324 		break;
325 	case AF_INET6:
326 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
327 						   flow_dissector,
328 						   target_container);
329 		if (dissector_uses_key(flow_dissector,
330 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
331 			struct flow_dissector_key_ipv6_addrs *ipv6;
332 
333 			ipv6 = skb_flow_dissector_target(flow_dissector,
334 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
335 							 target_container);
336 			ipv6->src = key->u.ipv6.src;
337 			ipv6->dst = key->u.ipv6.dst;
338 		}
339 		break;
340 	}
341 
342 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
343 		struct flow_dissector_key_keyid *keyid;
344 
345 		keyid = skb_flow_dissector_target(flow_dissector,
346 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
347 						  target_container);
348 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
349 	}
350 
351 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
352 		struct flow_dissector_key_ports *tp;
353 
354 		tp = skb_flow_dissector_target(flow_dissector,
355 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
356 					       target_container);
357 		tp->src = key->tp_src;
358 		tp->dst = key->tp_dst;
359 	}
360 
361 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
362 		struct flow_dissector_key_ip *ip;
363 
364 		ip = skb_flow_dissector_target(flow_dissector,
365 					       FLOW_DISSECTOR_KEY_ENC_IP,
366 					       target_container);
367 		ip->tos = key->tos;
368 		ip->ttl = key->ttl;
369 	}
370 
371 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
372 		struct flow_dissector_key_enc_opts *enc_opt;
373 
374 		enc_opt = skb_flow_dissector_target(flow_dissector,
375 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
376 						    target_container);
377 
378 		if (info->options_len) {
379 			enc_opt->len = info->options_len;
380 			ip_tunnel_info_opts_get(enc_opt->data, info);
381 			enc_opt->dst_opt_type = info->key.tun_flags &
382 						TUNNEL_OPTIONS_PRESENT;
383 		}
384 	}
385 }
386 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
387 
388 static enum flow_dissect_ret
389 __skb_flow_dissect_mpls(const struct sk_buff *skb,
390 			struct flow_dissector *flow_dissector,
391 			void *target_container, void *data, int nhoff, int hlen)
392 {
393 	struct flow_dissector_key_keyid *key_keyid;
394 	struct mpls_label *hdr, _hdr[2];
395 	u32 entry, label;
396 
397 	if (!dissector_uses_key(flow_dissector,
398 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
399 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
400 		return FLOW_DISSECT_RET_OUT_GOOD;
401 
402 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
403 				   hlen, &_hdr);
404 	if (!hdr)
405 		return FLOW_DISSECT_RET_OUT_BAD;
406 
407 	entry = ntohl(hdr[0].entry);
408 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
409 
410 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
411 		struct flow_dissector_key_mpls *key_mpls;
412 
413 		key_mpls = skb_flow_dissector_target(flow_dissector,
414 						     FLOW_DISSECTOR_KEY_MPLS,
415 						     target_container);
416 		key_mpls->mpls_label = label;
417 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
418 					>> MPLS_LS_TTL_SHIFT;
419 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
420 					>> MPLS_LS_TC_SHIFT;
421 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
422 					>> MPLS_LS_S_SHIFT;
423 	}
424 
425 	if (label == MPLS_LABEL_ENTROPY) {
426 		key_keyid = skb_flow_dissector_target(flow_dissector,
427 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
428 						      target_container);
429 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
430 	}
431 	return FLOW_DISSECT_RET_OUT_GOOD;
432 }
433 
434 static enum flow_dissect_ret
435 __skb_flow_dissect_arp(const struct sk_buff *skb,
436 		       struct flow_dissector *flow_dissector,
437 		       void *target_container, void *data, int nhoff, int hlen)
438 {
439 	struct flow_dissector_key_arp *key_arp;
440 	struct {
441 		unsigned char ar_sha[ETH_ALEN];
442 		unsigned char ar_sip[4];
443 		unsigned char ar_tha[ETH_ALEN];
444 		unsigned char ar_tip[4];
445 	} *arp_eth, _arp_eth;
446 	const struct arphdr *arp;
447 	struct arphdr _arp;
448 
449 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
450 		return FLOW_DISSECT_RET_OUT_GOOD;
451 
452 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
453 				   hlen, &_arp);
454 	if (!arp)
455 		return FLOW_DISSECT_RET_OUT_BAD;
456 
457 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
458 	    arp->ar_pro != htons(ETH_P_IP) ||
459 	    arp->ar_hln != ETH_ALEN ||
460 	    arp->ar_pln != 4 ||
461 	    (arp->ar_op != htons(ARPOP_REPLY) &&
462 	     arp->ar_op != htons(ARPOP_REQUEST)))
463 		return FLOW_DISSECT_RET_OUT_BAD;
464 
465 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
466 				       sizeof(_arp_eth), data,
467 				       hlen, &_arp_eth);
468 	if (!arp_eth)
469 		return FLOW_DISSECT_RET_OUT_BAD;
470 
471 	key_arp = skb_flow_dissector_target(flow_dissector,
472 					    FLOW_DISSECTOR_KEY_ARP,
473 					    target_container);
474 
475 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
476 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
477 
478 	/* Only store the lower byte of the opcode;
479 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
480 	 */
481 	key_arp->op = ntohs(arp->ar_op) & 0xff;
482 
483 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
484 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
485 
486 	return FLOW_DISSECT_RET_OUT_GOOD;
487 }
488 
489 static enum flow_dissect_ret
490 __skb_flow_dissect_gre(const struct sk_buff *skb,
491 		       struct flow_dissector_key_control *key_control,
492 		       struct flow_dissector *flow_dissector,
493 		       void *target_container, void *data,
494 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
495 		       unsigned int flags)
496 {
497 	struct flow_dissector_key_keyid *key_keyid;
498 	struct gre_base_hdr *hdr, _hdr;
499 	int offset = 0;
500 	u16 gre_ver;
501 
502 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
503 				   data, *p_hlen, &_hdr);
504 	if (!hdr)
505 		return FLOW_DISSECT_RET_OUT_BAD;
506 
507 	/* Only look inside GRE without routing */
508 	if (hdr->flags & GRE_ROUTING)
509 		return FLOW_DISSECT_RET_OUT_GOOD;
510 
511 	/* Only look inside GRE for version 0 and 1 */
512 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
513 	if (gre_ver > 1)
514 		return FLOW_DISSECT_RET_OUT_GOOD;
515 
516 	*p_proto = hdr->protocol;
517 	if (gre_ver) {
518 		/* Version1 must be PPTP, and check the flags */
519 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
520 			return FLOW_DISSECT_RET_OUT_GOOD;
521 	}
522 
523 	offset += sizeof(struct gre_base_hdr);
524 
525 	if (hdr->flags & GRE_CSUM)
526 		offset += FIELD_SIZEOF(struct gre_full_hdr, csum) +
527 			  FIELD_SIZEOF(struct gre_full_hdr, reserved1);
528 
529 	if (hdr->flags & GRE_KEY) {
530 		const __be32 *keyid;
531 		__be32 _keyid;
532 
533 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
534 					     sizeof(_keyid),
535 					     data, *p_hlen, &_keyid);
536 		if (!keyid)
537 			return FLOW_DISSECT_RET_OUT_BAD;
538 
539 		if (dissector_uses_key(flow_dissector,
540 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
541 			key_keyid = skb_flow_dissector_target(flow_dissector,
542 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
543 							      target_container);
544 			if (gre_ver == 0)
545 				key_keyid->keyid = *keyid;
546 			else
547 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
548 		}
549 		offset += FIELD_SIZEOF(struct gre_full_hdr, key);
550 	}
551 
552 	if (hdr->flags & GRE_SEQ)
553 		offset += FIELD_SIZEOF(struct pptp_gre_header, seq);
554 
555 	if (gre_ver == 0) {
556 		if (*p_proto == htons(ETH_P_TEB)) {
557 			const struct ethhdr *eth;
558 			struct ethhdr _eth;
559 
560 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
561 						   sizeof(_eth),
562 						   data, *p_hlen, &_eth);
563 			if (!eth)
564 				return FLOW_DISSECT_RET_OUT_BAD;
565 			*p_proto = eth->h_proto;
566 			offset += sizeof(*eth);
567 
568 			/* Cap headers that we access via pointers at the
569 			 * end of the Ethernet header as our maximum alignment
570 			 * at that point is only 2 bytes.
571 			 */
572 			if (NET_IP_ALIGN)
573 				*p_hlen = *p_nhoff + offset;
574 		}
575 	} else { /* version 1, must be PPTP */
576 		u8 _ppp_hdr[PPP_HDRLEN];
577 		u8 *ppp_hdr;
578 
579 		if (hdr->flags & GRE_ACK)
580 			offset += FIELD_SIZEOF(struct pptp_gre_header, ack);
581 
582 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
583 					       sizeof(_ppp_hdr),
584 					       data, *p_hlen, _ppp_hdr);
585 		if (!ppp_hdr)
586 			return FLOW_DISSECT_RET_OUT_BAD;
587 
588 		switch (PPP_PROTOCOL(ppp_hdr)) {
589 		case PPP_IP:
590 			*p_proto = htons(ETH_P_IP);
591 			break;
592 		case PPP_IPV6:
593 			*p_proto = htons(ETH_P_IPV6);
594 			break;
595 		default:
596 			/* Could probably catch some more like MPLS */
597 			break;
598 		}
599 
600 		offset += PPP_HDRLEN;
601 	}
602 
603 	*p_nhoff += offset;
604 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
605 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
606 		return FLOW_DISSECT_RET_OUT_GOOD;
607 
608 	return FLOW_DISSECT_RET_PROTO_AGAIN;
609 }
610 
611 /**
612  * __skb_flow_dissect_batadv() - dissect batman-adv header
613  * @skb: sk_buff to with the batman-adv header
614  * @key_control: flow dissectors control key
615  * @data: raw buffer pointer to the packet, if NULL use skb->data
616  * @p_proto: pointer used to update the protocol to process next
617  * @p_nhoff: pointer used to update inner network header offset
618  * @hlen: packet header length
619  * @flags: any combination of FLOW_DISSECTOR_F_*
620  *
621  * ETH_P_BATMAN packets are tried to be dissected. Only
622  * &struct batadv_unicast packets are actually processed because they contain an
623  * inner ethernet header and are usually followed by actual network header. This
624  * allows the flow dissector to continue processing the packet.
625  *
626  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
627  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
628  *  otherwise FLOW_DISSECT_RET_OUT_BAD
629  */
630 static enum flow_dissect_ret
631 __skb_flow_dissect_batadv(const struct sk_buff *skb,
632 			  struct flow_dissector_key_control *key_control,
633 			  void *data, __be16 *p_proto, int *p_nhoff, int hlen,
634 			  unsigned int flags)
635 {
636 	struct {
637 		struct batadv_unicast_packet batadv_unicast;
638 		struct ethhdr eth;
639 	} *hdr, _hdr;
640 
641 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
642 				   &_hdr);
643 	if (!hdr)
644 		return FLOW_DISSECT_RET_OUT_BAD;
645 
646 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
647 		return FLOW_DISSECT_RET_OUT_BAD;
648 
649 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
650 		return FLOW_DISSECT_RET_OUT_BAD;
651 
652 	*p_proto = hdr->eth.h_proto;
653 	*p_nhoff += sizeof(*hdr);
654 
655 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
656 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
657 		return FLOW_DISSECT_RET_OUT_GOOD;
658 
659 	return FLOW_DISSECT_RET_PROTO_AGAIN;
660 }
661 
662 static void
663 __skb_flow_dissect_tcp(const struct sk_buff *skb,
664 		       struct flow_dissector *flow_dissector,
665 		       void *target_container, void *data, int thoff, int hlen)
666 {
667 	struct flow_dissector_key_tcp *key_tcp;
668 	struct tcphdr *th, _th;
669 
670 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
671 		return;
672 
673 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
674 	if (!th)
675 		return;
676 
677 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
678 		return;
679 
680 	key_tcp = skb_flow_dissector_target(flow_dissector,
681 					    FLOW_DISSECTOR_KEY_TCP,
682 					    target_container);
683 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
684 }
685 
686 static void
687 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
688 			struct flow_dissector *flow_dissector,
689 			void *target_container, void *data, const struct iphdr *iph)
690 {
691 	struct flow_dissector_key_ip *key_ip;
692 
693 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
694 		return;
695 
696 	key_ip = skb_flow_dissector_target(flow_dissector,
697 					   FLOW_DISSECTOR_KEY_IP,
698 					   target_container);
699 	key_ip->tos = iph->tos;
700 	key_ip->ttl = iph->ttl;
701 }
702 
703 static void
704 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
705 			struct flow_dissector *flow_dissector,
706 			void *target_container, void *data, const struct ipv6hdr *iph)
707 {
708 	struct flow_dissector_key_ip *key_ip;
709 
710 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
711 		return;
712 
713 	key_ip = skb_flow_dissector_target(flow_dissector,
714 					   FLOW_DISSECTOR_KEY_IP,
715 					   target_container);
716 	key_ip->tos = ipv6_get_dsfield(iph);
717 	key_ip->ttl = iph->hop_limit;
718 }
719 
720 /* Maximum number of protocol headers that can be parsed in
721  * __skb_flow_dissect
722  */
723 #define MAX_FLOW_DISSECT_HDRS	15
724 
725 static bool skb_flow_dissect_allowed(int *num_hdrs)
726 {
727 	++*num_hdrs;
728 
729 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
730 }
731 
732 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
733 				     struct flow_dissector *flow_dissector,
734 				     void *target_container)
735 {
736 	struct flow_dissector_key_control *key_control;
737 	struct flow_dissector_key_basic *key_basic;
738 	struct flow_dissector_key_addrs *key_addrs;
739 	struct flow_dissector_key_ports *key_ports;
740 
741 	key_control = skb_flow_dissector_target(flow_dissector,
742 						FLOW_DISSECTOR_KEY_CONTROL,
743 						target_container);
744 	key_control->thoff = flow_keys->thoff;
745 	if (flow_keys->is_frag)
746 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
747 	if (flow_keys->is_first_frag)
748 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
749 	if (flow_keys->is_encap)
750 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
751 
752 	key_basic = skb_flow_dissector_target(flow_dissector,
753 					      FLOW_DISSECTOR_KEY_BASIC,
754 					      target_container);
755 	key_basic->n_proto = flow_keys->n_proto;
756 	key_basic->ip_proto = flow_keys->ip_proto;
757 
758 	if (flow_keys->addr_proto == ETH_P_IP &&
759 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
760 		key_addrs = skb_flow_dissector_target(flow_dissector,
761 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
762 						      target_container);
763 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
764 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
765 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
766 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
767 		   dissector_uses_key(flow_dissector,
768 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
769 		key_addrs = skb_flow_dissector_target(flow_dissector,
770 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
771 						      target_container);
772 		memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src,
773 		       sizeof(key_addrs->v6addrs));
774 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
775 	}
776 
777 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) {
778 		key_ports = skb_flow_dissector_target(flow_dissector,
779 						      FLOW_DISSECTOR_KEY_PORTS,
780 						      target_container);
781 		key_ports->src = flow_keys->sport;
782 		key_ports->dst = flow_keys->dport;
783 	}
784 }
785 
786 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
787 		      __be16 proto, int nhoff, int hlen)
788 {
789 	struct bpf_flow_keys *flow_keys = ctx->flow_keys;
790 	u32 result;
791 
792 	/* Pass parameters to the BPF program */
793 	memset(flow_keys, 0, sizeof(*flow_keys));
794 	flow_keys->n_proto = proto;
795 	flow_keys->nhoff = nhoff;
796 	flow_keys->thoff = flow_keys->nhoff;
797 
798 	preempt_disable();
799 	result = BPF_PROG_RUN(prog, ctx);
800 	preempt_enable();
801 
802 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
803 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
804 				   flow_keys->nhoff, hlen);
805 
806 	return result == BPF_OK;
807 }
808 
809 /**
810  * __skb_flow_dissect - extract the flow_keys struct and return it
811  * @net: associated network namespace, derived from @skb if NULL
812  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
813  * @flow_dissector: list of keys to dissect
814  * @target_container: target structure to put dissected values into
815  * @data: raw buffer pointer to the packet, if NULL use skb->data
816  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
817  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
818  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
819  * @flags: flags that control the dissection process, e.g.
820  *         FLOW_DISSECTOR_F_STOP_AT_ENCAP.
821  *
822  * The function will try to retrieve individual keys into target specified
823  * by flow_dissector from either the skbuff or a raw buffer specified by the
824  * rest parameters.
825  *
826  * Caller must take care of zeroing target container memory.
827  */
828 bool __skb_flow_dissect(const struct net *net,
829 			const struct sk_buff *skb,
830 			struct flow_dissector *flow_dissector,
831 			void *target_container,
832 			void *data, __be16 proto, int nhoff, int hlen,
833 			unsigned int flags)
834 {
835 	struct flow_dissector_key_control *key_control;
836 	struct flow_dissector_key_basic *key_basic;
837 	struct flow_dissector_key_addrs *key_addrs;
838 	struct flow_dissector_key_ports *key_ports;
839 	struct flow_dissector_key_icmp *key_icmp;
840 	struct flow_dissector_key_tags *key_tags;
841 	struct flow_dissector_key_vlan *key_vlan;
842 	struct bpf_prog *attached = NULL;
843 	enum flow_dissect_ret fdret;
844 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
845 	int num_hdrs = 0;
846 	u8 ip_proto = 0;
847 	bool ret;
848 
849 	if (!data) {
850 		data = skb->data;
851 		proto = skb_vlan_tag_present(skb) ?
852 			 skb->vlan_proto : skb->protocol;
853 		nhoff = skb_network_offset(skb);
854 		hlen = skb_headlen(skb);
855 #if IS_ENABLED(CONFIG_NET_DSA)
856 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
857 			const struct dsa_device_ops *ops;
858 			int offset;
859 
860 			ops = skb->dev->dsa_ptr->tag_ops;
861 			if (ops->flow_dissect &&
862 			    !ops->flow_dissect(skb, &proto, &offset)) {
863 				hlen -= offset;
864 				nhoff += offset;
865 			}
866 		}
867 #endif
868 	}
869 
870 	/* It is ensured by skb_flow_dissector_init() that control key will
871 	 * be always present.
872 	 */
873 	key_control = skb_flow_dissector_target(flow_dissector,
874 						FLOW_DISSECTOR_KEY_CONTROL,
875 						target_container);
876 
877 	/* It is ensured by skb_flow_dissector_init() that basic key will
878 	 * be always present.
879 	 */
880 	key_basic = skb_flow_dissector_target(flow_dissector,
881 					      FLOW_DISSECTOR_KEY_BASIC,
882 					      target_container);
883 
884 	if (skb) {
885 		if (!net) {
886 			if (skb->dev)
887 				net = dev_net(skb->dev);
888 			else if (skb->sk)
889 				net = sock_net(skb->sk);
890 		}
891 	}
892 
893 	WARN_ON_ONCE(!net);
894 	if (net) {
895 		rcu_read_lock();
896 		attached = rcu_dereference(net->flow_dissector_prog);
897 
898 		if (attached) {
899 			struct bpf_flow_keys flow_keys;
900 			struct bpf_flow_dissector ctx = {
901 				.flow_keys = &flow_keys,
902 				.data = data,
903 				.data_end = data + hlen,
904 			};
905 			__be16 n_proto = proto;
906 
907 			if (skb) {
908 				ctx.skb = skb;
909 				/* we can't use 'proto' in the skb case
910 				 * because it might be set to skb->vlan_proto
911 				 * which has been pulled from the data
912 				 */
913 				n_proto = skb->protocol;
914 			}
915 
916 			ret = bpf_flow_dissect(attached, &ctx, n_proto, nhoff,
917 					       hlen);
918 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
919 						 target_container);
920 			rcu_read_unlock();
921 			return ret;
922 		}
923 		rcu_read_unlock();
924 	}
925 
926 	if (dissector_uses_key(flow_dissector,
927 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
928 		struct ethhdr *eth = eth_hdr(skb);
929 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
930 
931 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
932 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
933 							  target_container);
934 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
935 	}
936 
937 proto_again:
938 	fdret = FLOW_DISSECT_RET_CONTINUE;
939 
940 	switch (proto) {
941 	case htons(ETH_P_IP): {
942 		const struct iphdr *iph;
943 		struct iphdr _iph;
944 
945 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
946 		if (!iph || iph->ihl < 5) {
947 			fdret = FLOW_DISSECT_RET_OUT_BAD;
948 			break;
949 		}
950 
951 		nhoff += iph->ihl * 4;
952 
953 		ip_proto = iph->protocol;
954 
955 		if (dissector_uses_key(flow_dissector,
956 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
957 			key_addrs = skb_flow_dissector_target(flow_dissector,
958 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
959 							      target_container);
960 
961 			memcpy(&key_addrs->v4addrs, &iph->saddr,
962 			       sizeof(key_addrs->v4addrs));
963 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
964 		}
965 
966 		if (ip_is_fragment(iph)) {
967 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
968 
969 			if (iph->frag_off & htons(IP_OFFSET)) {
970 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
971 				break;
972 			} else {
973 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
974 				if (!(flags &
975 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
976 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
977 					break;
978 				}
979 			}
980 		}
981 
982 		__skb_flow_dissect_ipv4(skb, flow_dissector,
983 					target_container, data, iph);
984 
985 		break;
986 	}
987 	case htons(ETH_P_IPV6): {
988 		const struct ipv6hdr *iph;
989 		struct ipv6hdr _iph;
990 
991 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
992 		if (!iph) {
993 			fdret = FLOW_DISSECT_RET_OUT_BAD;
994 			break;
995 		}
996 
997 		ip_proto = iph->nexthdr;
998 		nhoff += sizeof(struct ipv6hdr);
999 
1000 		if (dissector_uses_key(flow_dissector,
1001 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1002 			key_addrs = skb_flow_dissector_target(flow_dissector,
1003 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1004 							      target_container);
1005 
1006 			memcpy(&key_addrs->v6addrs, &iph->saddr,
1007 			       sizeof(key_addrs->v6addrs));
1008 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1009 		}
1010 
1011 		if ((dissector_uses_key(flow_dissector,
1012 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1013 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1014 		    ip6_flowlabel(iph)) {
1015 			__be32 flow_label = ip6_flowlabel(iph);
1016 
1017 			if (dissector_uses_key(flow_dissector,
1018 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1019 				key_tags = skb_flow_dissector_target(flow_dissector,
1020 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
1021 								     target_container);
1022 				key_tags->flow_label = ntohl(flow_label);
1023 			}
1024 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1025 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1026 				break;
1027 			}
1028 		}
1029 
1030 		__skb_flow_dissect_ipv6(skb, flow_dissector,
1031 					target_container, data, iph);
1032 
1033 		break;
1034 	}
1035 	case htons(ETH_P_8021AD):
1036 	case htons(ETH_P_8021Q): {
1037 		const struct vlan_hdr *vlan = NULL;
1038 		struct vlan_hdr _vlan;
1039 		__be16 saved_vlan_tpid = proto;
1040 
1041 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1042 		    skb && skb_vlan_tag_present(skb)) {
1043 			proto = skb->protocol;
1044 		} else {
1045 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1046 						    data, hlen, &_vlan);
1047 			if (!vlan) {
1048 				fdret = FLOW_DISSECT_RET_OUT_BAD;
1049 				break;
1050 			}
1051 
1052 			proto = vlan->h_vlan_encapsulated_proto;
1053 			nhoff += sizeof(*vlan);
1054 		}
1055 
1056 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1057 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1058 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1059 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1060 		} else {
1061 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1062 			break;
1063 		}
1064 
1065 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1066 			key_vlan = skb_flow_dissector_target(flow_dissector,
1067 							     dissector_vlan,
1068 							     target_container);
1069 
1070 			if (!vlan) {
1071 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1072 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1073 			} else {
1074 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1075 					VLAN_VID_MASK;
1076 				key_vlan->vlan_priority =
1077 					(ntohs(vlan->h_vlan_TCI) &
1078 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1079 			}
1080 			key_vlan->vlan_tpid = saved_vlan_tpid;
1081 		}
1082 
1083 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1084 		break;
1085 	}
1086 	case htons(ETH_P_PPP_SES): {
1087 		struct {
1088 			struct pppoe_hdr hdr;
1089 			__be16 proto;
1090 		} *hdr, _hdr;
1091 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1092 		if (!hdr) {
1093 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1094 			break;
1095 		}
1096 
1097 		proto = hdr->proto;
1098 		nhoff += PPPOE_SES_HLEN;
1099 		switch (proto) {
1100 		case htons(PPP_IP):
1101 			proto = htons(ETH_P_IP);
1102 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1103 			break;
1104 		case htons(PPP_IPV6):
1105 			proto = htons(ETH_P_IPV6);
1106 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1107 			break;
1108 		default:
1109 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1110 			break;
1111 		}
1112 		break;
1113 	}
1114 	case htons(ETH_P_TIPC): {
1115 		struct tipc_basic_hdr *hdr, _hdr;
1116 
1117 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1118 					   data, hlen, &_hdr);
1119 		if (!hdr) {
1120 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1121 			break;
1122 		}
1123 
1124 		if (dissector_uses_key(flow_dissector,
1125 				       FLOW_DISSECTOR_KEY_TIPC)) {
1126 			key_addrs = skb_flow_dissector_target(flow_dissector,
1127 							      FLOW_DISSECTOR_KEY_TIPC,
1128 							      target_container);
1129 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1130 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1131 		}
1132 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1133 		break;
1134 	}
1135 
1136 	case htons(ETH_P_MPLS_UC):
1137 	case htons(ETH_P_MPLS_MC):
1138 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1139 						target_container, data,
1140 						nhoff, hlen);
1141 		break;
1142 	case htons(ETH_P_FCOE):
1143 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1144 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1145 			break;
1146 		}
1147 
1148 		nhoff += FCOE_HEADER_LEN;
1149 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1150 		break;
1151 
1152 	case htons(ETH_P_ARP):
1153 	case htons(ETH_P_RARP):
1154 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1155 					       target_container, data,
1156 					       nhoff, hlen);
1157 		break;
1158 
1159 	case htons(ETH_P_BATMAN):
1160 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1161 						  &proto, &nhoff, hlen, flags);
1162 		break;
1163 
1164 	default:
1165 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1166 		break;
1167 	}
1168 
1169 	/* Process result of proto processing */
1170 	switch (fdret) {
1171 	case FLOW_DISSECT_RET_OUT_GOOD:
1172 		goto out_good;
1173 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1174 		if (skb_flow_dissect_allowed(&num_hdrs))
1175 			goto proto_again;
1176 		goto out_good;
1177 	case FLOW_DISSECT_RET_CONTINUE:
1178 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1179 		break;
1180 	case FLOW_DISSECT_RET_OUT_BAD:
1181 	default:
1182 		goto out_bad;
1183 	}
1184 
1185 ip_proto_again:
1186 	fdret = FLOW_DISSECT_RET_CONTINUE;
1187 
1188 	switch (ip_proto) {
1189 	case IPPROTO_GRE:
1190 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1191 					       target_container, data,
1192 					       &proto, &nhoff, &hlen, flags);
1193 		break;
1194 
1195 	case NEXTHDR_HOP:
1196 	case NEXTHDR_ROUTING:
1197 	case NEXTHDR_DEST: {
1198 		u8 _opthdr[2], *opthdr;
1199 
1200 		if (proto != htons(ETH_P_IPV6))
1201 			break;
1202 
1203 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1204 					      data, hlen, &_opthdr);
1205 		if (!opthdr) {
1206 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1207 			break;
1208 		}
1209 
1210 		ip_proto = opthdr[0];
1211 		nhoff += (opthdr[1] + 1) << 3;
1212 
1213 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1214 		break;
1215 	}
1216 	case NEXTHDR_FRAGMENT: {
1217 		struct frag_hdr _fh, *fh;
1218 
1219 		if (proto != htons(ETH_P_IPV6))
1220 			break;
1221 
1222 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1223 					  data, hlen, &_fh);
1224 
1225 		if (!fh) {
1226 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1227 			break;
1228 		}
1229 
1230 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1231 
1232 		nhoff += sizeof(_fh);
1233 		ip_proto = fh->nexthdr;
1234 
1235 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1236 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1237 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1238 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1239 				break;
1240 			}
1241 		}
1242 
1243 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1244 		break;
1245 	}
1246 	case IPPROTO_IPIP:
1247 		proto = htons(ETH_P_IP);
1248 
1249 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1250 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1251 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1252 			break;
1253 		}
1254 
1255 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1256 		break;
1257 
1258 	case IPPROTO_IPV6:
1259 		proto = htons(ETH_P_IPV6);
1260 
1261 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1262 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1263 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1264 			break;
1265 		}
1266 
1267 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1268 		break;
1269 
1270 
1271 	case IPPROTO_MPLS:
1272 		proto = htons(ETH_P_MPLS_UC);
1273 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1274 		break;
1275 
1276 	case IPPROTO_TCP:
1277 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1278 				       data, nhoff, hlen);
1279 		break;
1280 
1281 	default:
1282 		break;
1283 	}
1284 
1285 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) &&
1286 	    !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) {
1287 		key_ports = skb_flow_dissector_target(flow_dissector,
1288 						      FLOW_DISSECTOR_KEY_PORTS,
1289 						      target_container);
1290 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1291 							data, hlen);
1292 	}
1293 
1294 	if (dissector_uses_key(flow_dissector,
1295 			       FLOW_DISSECTOR_KEY_ICMP)) {
1296 		key_icmp = skb_flow_dissector_target(flow_dissector,
1297 						     FLOW_DISSECTOR_KEY_ICMP,
1298 						     target_container);
1299 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1300 	}
1301 
1302 	/* Process result of IP proto processing */
1303 	switch (fdret) {
1304 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1305 		if (skb_flow_dissect_allowed(&num_hdrs))
1306 			goto proto_again;
1307 		break;
1308 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1309 		if (skb_flow_dissect_allowed(&num_hdrs))
1310 			goto ip_proto_again;
1311 		break;
1312 	case FLOW_DISSECT_RET_OUT_GOOD:
1313 	case FLOW_DISSECT_RET_CONTINUE:
1314 		break;
1315 	case FLOW_DISSECT_RET_OUT_BAD:
1316 	default:
1317 		goto out_bad;
1318 	}
1319 
1320 out_good:
1321 	ret = true;
1322 
1323 out:
1324 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1325 	key_basic->n_proto = proto;
1326 	key_basic->ip_proto = ip_proto;
1327 
1328 	return ret;
1329 
1330 out_bad:
1331 	ret = false;
1332 	goto out;
1333 }
1334 EXPORT_SYMBOL(__skb_flow_dissect);
1335 
1336 static u32 hashrnd __read_mostly;
1337 static __always_inline void __flow_hash_secret_init(void)
1338 {
1339 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1340 }
1341 
1342 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1343 					     u32 keyval)
1344 {
1345 	return jhash2(words, length, keyval);
1346 }
1347 
1348 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1349 {
1350 	const void *p = flow;
1351 
1352 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1353 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1354 }
1355 
1356 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1357 {
1358 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1359 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1360 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1361 		     sizeof(*flow) - sizeof(flow->addrs));
1362 
1363 	switch (flow->control.addr_type) {
1364 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1365 		diff -= sizeof(flow->addrs.v4addrs);
1366 		break;
1367 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1368 		diff -= sizeof(flow->addrs.v6addrs);
1369 		break;
1370 	case FLOW_DISSECTOR_KEY_TIPC:
1371 		diff -= sizeof(flow->addrs.tipckey);
1372 		break;
1373 	}
1374 	return (sizeof(*flow) - diff) / sizeof(u32);
1375 }
1376 
1377 __be32 flow_get_u32_src(const struct flow_keys *flow)
1378 {
1379 	switch (flow->control.addr_type) {
1380 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1381 		return flow->addrs.v4addrs.src;
1382 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1383 		return (__force __be32)ipv6_addr_hash(
1384 			&flow->addrs.v6addrs.src);
1385 	case FLOW_DISSECTOR_KEY_TIPC:
1386 		return flow->addrs.tipckey.key;
1387 	default:
1388 		return 0;
1389 	}
1390 }
1391 EXPORT_SYMBOL(flow_get_u32_src);
1392 
1393 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1394 {
1395 	switch (flow->control.addr_type) {
1396 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1397 		return flow->addrs.v4addrs.dst;
1398 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1399 		return (__force __be32)ipv6_addr_hash(
1400 			&flow->addrs.v6addrs.dst);
1401 	default:
1402 		return 0;
1403 	}
1404 }
1405 EXPORT_SYMBOL(flow_get_u32_dst);
1406 
1407 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1408 {
1409 	int addr_diff, i;
1410 
1411 	switch (keys->control.addr_type) {
1412 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1413 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1414 			    (__force u32)keys->addrs.v4addrs.src;
1415 		if ((addr_diff < 0) ||
1416 		    (addr_diff == 0 &&
1417 		     ((__force u16)keys->ports.dst <
1418 		      (__force u16)keys->ports.src))) {
1419 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1420 			swap(keys->ports.src, keys->ports.dst);
1421 		}
1422 		break;
1423 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1424 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1425 				   &keys->addrs.v6addrs.src,
1426 				   sizeof(keys->addrs.v6addrs.dst));
1427 		if ((addr_diff < 0) ||
1428 		    (addr_diff == 0 &&
1429 		     ((__force u16)keys->ports.dst <
1430 		      (__force u16)keys->ports.src))) {
1431 			for (i = 0; i < 4; i++)
1432 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1433 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1434 			swap(keys->ports.src, keys->ports.dst);
1435 		}
1436 		break;
1437 	}
1438 }
1439 
1440 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1441 {
1442 	u32 hash;
1443 
1444 	__flow_hash_consistentify(keys);
1445 
1446 	hash = __flow_hash_words(flow_keys_hash_start(keys),
1447 				 flow_keys_hash_length(keys), keyval);
1448 	if (!hash)
1449 		hash = 1;
1450 
1451 	return hash;
1452 }
1453 
1454 u32 flow_hash_from_keys(struct flow_keys *keys)
1455 {
1456 	__flow_hash_secret_init();
1457 	return __flow_hash_from_keys(keys, hashrnd);
1458 }
1459 EXPORT_SYMBOL(flow_hash_from_keys);
1460 
1461 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1462 				  struct flow_keys *keys, u32 keyval)
1463 {
1464 	skb_flow_dissect_flow_keys(skb, keys,
1465 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1466 
1467 	return __flow_hash_from_keys(keys, keyval);
1468 }
1469 
1470 struct _flow_keys_digest_data {
1471 	__be16	n_proto;
1472 	u8	ip_proto;
1473 	u8	padding;
1474 	__be32	ports;
1475 	__be32	src;
1476 	__be32	dst;
1477 };
1478 
1479 void make_flow_keys_digest(struct flow_keys_digest *digest,
1480 			   const struct flow_keys *flow)
1481 {
1482 	struct _flow_keys_digest_data *data =
1483 	    (struct _flow_keys_digest_data *)digest;
1484 
1485 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1486 
1487 	memset(digest, 0, sizeof(*digest));
1488 
1489 	data->n_proto = flow->basic.n_proto;
1490 	data->ip_proto = flow->basic.ip_proto;
1491 	data->ports = flow->ports.ports;
1492 	data->src = flow->addrs.v4addrs.src;
1493 	data->dst = flow->addrs.v4addrs.dst;
1494 }
1495 EXPORT_SYMBOL(make_flow_keys_digest);
1496 
1497 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1498 
1499 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1500 {
1501 	struct flow_keys keys;
1502 
1503 	__flow_hash_secret_init();
1504 
1505 	memset(&keys, 0, sizeof(keys));
1506 	__skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1507 			   &keys, NULL, 0, 0, 0,
1508 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1509 
1510 	return __flow_hash_from_keys(&keys, hashrnd);
1511 }
1512 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1513 
1514 /**
1515  * __skb_get_hash: calculate a flow hash
1516  * @skb: sk_buff to calculate flow hash from
1517  *
1518  * This function calculates a flow hash based on src/dst addresses
1519  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1520  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1521  * if hash is a canonical 4-tuple hash over transport ports.
1522  */
1523 void __skb_get_hash(struct sk_buff *skb)
1524 {
1525 	struct flow_keys keys;
1526 	u32 hash;
1527 
1528 	__flow_hash_secret_init();
1529 
1530 	hash = ___skb_get_hash(skb, &keys, hashrnd);
1531 
1532 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1533 }
1534 EXPORT_SYMBOL(__skb_get_hash);
1535 
1536 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1537 {
1538 	struct flow_keys keys;
1539 
1540 	return ___skb_get_hash(skb, &keys, perturb);
1541 }
1542 EXPORT_SYMBOL(skb_get_hash_perturb);
1543 
1544 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1545 		   const struct flow_keys_basic *keys, int hlen)
1546 {
1547 	u32 poff = keys->control.thoff;
1548 
1549 	/* skip L4 headers for fragments after the first */
1550 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1551 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1552 		return poff;
1553 
1554 	switch (keys->basic.ip_proto) {
1555 	case IPPROTO_TCP: {
1556 		/* access doff as u8 to avoid unaligned access */
1557 		const u8 *doff;
1558 		u8 _doff;
1559 
1560 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1561 					    data, hlen, &_doff);
1562 		if (!doff)
1563 			return poff;
1564 
1565 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1566 		break;
1567 	}
1568 	case IPPROTO_UDP:
1569 	case IPPROTO_UDPLITE:
1570 		poff += sizeof(struct udphdr);
1571 		break;
1572 	/* For the rest, we do not really care about header
1573 	 * extensions at this point for now.
1574 	 */
1575 	case IPPROTO_ICMP:
1576 		poff += sizeof(struct icmphdr);
1577 		break;
1578 	case IPPROTO_ICMPV6:
1579 		poff += sizeof(struct icmp6hdr);
1580 		break;
1581 	case IPPROTO_IGMP:
1582 		poff += sizeof(struct igmphdr);
1583 		break;
1584 	case IPPROTO_DCCP:
1585 		poff += sizeof(struct dccp_hdr);
1586 		break;
1587 	case IPPROTO_SCTP:
1588 		poff += sizeof(struct sctphdr);
1589 		break;
1590 	}
1591 
1592 	return poff;
1593 }
1594 
1595 /**
1596  * skb_get_poff - get the offset to the payload
1597  * @skb: sk_buff to get the payload offset from
1598  *
1599  * The function will get the offset to the payload as far as it could
1600  * be dissected.  The main user is currently BPF, so that we can dynamically
1601  * truncate packets without needing to push actual payload to the user
1602  * space and can analyze headers only, instead.
1603  */
1604 u32 skb_get_poff(const struct sk_buff *skb)
1605 {
1606 	struct flow_keys_basic keys;
1607 
1608 	if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1609 					      NULL, 0, 0, 0, 0))
1610 		return 0;
1611 
1612 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1613 }
1614 
1615 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1616 {
1617 	memset(keys, 0, sizeof(*keys));
1618 
1619 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1620 	    sizeof(keys->addrs.v6addrs.src));
1621 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1622 	    sizeof(keys->addrs.v6addrs.dst));
1623 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1624 	keys->ports.src = fl6->fl6_sport;
1625 	keys->ports.dst = fl6->fl6_dport;
1626 	keys->keyid.keyid = fl6->fl6_gre_key;
1627 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1628 	keys->basic.ip_proto = fl6->flowi6_proto;
1629 
1630 	return flow_hash_from_keys(keys);
1631 }
1632 EXPORT_SYMBOL(__get_hash_from_flowi6);
1633 
1634 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1635 	{
1636 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1637 		.offset = offsetof(struct flow_keys, control),
1638 	},
1639 	{
1640 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1641 		.offset = offsetof(struct flow_keys, basic),
1642 	},
1643 	{
1644 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1645 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1646 	},
1647 	{
1648 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1649 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1650 	},
1651 	{
1652 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1653 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1654 	},
1655 	{
1656 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1657 		.offset = offsetof(struct flow_keys, ports),
1658 	},
1659 	{
1660 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1661 		.offset = offsetof(struct flow_keys, vlan),
1662 	},
1663 	{
1664 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1665 		.offset = offsetof(struct flow_keys, tags),
1666 	},
1667 	{
1668 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1669 		.offset = offsetof(struct flow_keys, keyid),
1670 	},
1671 };
1672 
1673 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1674 	{
1675 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1676 		.offset = offsetof(struct flow_keys, control),
1677 	},
1678 	{
1679 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1680 		.offset = offsetof(struct flow_keys, basic),
1681 	},
1682 	{
1683 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1684 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1685 	},
1686 	{
1687 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1688 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1689 	},
1690 	{
1691 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1692 		.offset = offsetof(struct flow_keys, ports),
1693 	},
1694 };
1695 
1696 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1697 	{
1698 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1699 		.offset = offsetof(struct flow_keys, control),
1700 	},
1701 	{
1702 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1703 		.offset = offsetof(struct flow_keys, basic),
1704 	},
1705 };
1706 
1707 struct flow_dissector flow_keys_dissector __read_mostly;
1708 EXPORT_SYMBOL(flow_keys_dissector);
1709 
1710 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1711 EXPORT_SYMBOL(flow_keys_basic_dissector);
1712 
1713 static int __init init_default_flow_dissectors(void)
1714 {
1715 	skb_flow_dissector_init(&flow_keys_dissector,
1716 				flow_keys_dissector_keys,
1717 				ARRAY_SIZE(flow_keys_dissector_keys));
1718 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1719 				flow_keys_dissector_symmetric_keys,
1720 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1721 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1722 				flow_keys_basic_dissector_keys,
1723 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1724 	return 0;
1725 }
1726 
1727 core_initcall(init_default_flow_dissectors);
1728