xref: /linux/net/core/flow_dissector.c (revision af8e51644a70f612974a6e767fa7d896d3c23f88)
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 <linux/filter.h>
9 #include <net/dsa.h>
10 #include <net/dst_metadata.h>
11 #include <net/ip.h>
12 #include <net/ipv6.h>
13 #include <net/gre.h>
14 #include <net/pptp.h>
15 #include <net/tipc.h>
16 #include <linux/igmp.h>
17 #include <linux/icmp.h>
18 #include <linux/sctp.h>
19 #include <linux/dccp.h>
20 #include <linux/if_tunnel.h>
21 #include <linux/if_pppox.h>
22 #include <linux/ppp_defs.h>
23 #include <linux/stddef.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_hsr.h>
26 #include <linux/mpls.h>
27 #include <linux/tcp.h>
28 #include <linux/ptp_classify.h>
29 #include <net/flow_dissector.h>
30 #include <net/pkt_cls.h>
31 #include <scsi/fc/fc_fcoe.h>
32 #include <uapi/linux/batadv_packet.h>
33 #include <linux/bpf.h>
34 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
35 #include <net/netfilter/nf_conntrack_core.h>
36 #include <net/netfilter/nf_conntrack_labels.h>
37 #endif
38 #include <linux/bpf-netns.h>
39 
40 static void dissector_set_key(struct flow_dissector *flow_dissector,
41 			      enum flow_dissector_key_id key_id)
42 {
43 	flow_dissector->used_keys |= (1ULL << key_id);
44 }
45 
46 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
47 			     const struct flow_dissector_key *key,
48 			     unsigned int key_count)
49 {
50 	unsigned int i;
51 
52 	memset(flow_dissector, 0, sizeof(*flow_dissector));
53 
54 	for (i = 0; i < key_count; i++, key++) {
55 		/* User should make sure that every key target offset is within
56 		 * boundaries of unsigned short.
57 		 */
58 		BUG_ON(key->offset > USHRT_MAX);
59 		BUG_ON(dissector_uses_key(flow_dissector,
60 					  key->key_id));
61 
62 		dissector_set_key(flow_dissector, key->key_id);
63 		flow_dissector->offset[key->key_id] = key->offset;
64 	}
65 
66 	/* Ensure that the dissector always includes control and basic key.
67 	 * That way we are able to avoid handling lack of these in fast path.
68 	 */
69 	BUG_ON(!dissector_uses_key(flow_dissector,
70 				   FLOW_DISSECTOR_KEY_CONTROL));
71 	BUG_ON(!dissector_uses_key(flow_dissector,
72 				   FLOW_DISSECTOR_KEY_BASIC));
73 }
74 EXPORT_SYMBOL(skb_flow_dissector_init);
75 
76 #ifdef CONFIG_BPF_SYSCALL
77 int flow_dissector_bpf_prog_attach_check(struct net *net,
78 					 struct bpf_prog *prog)
79 {
80 	enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
81 
82 	if (net == &init_net) {
83 		/* BPF flow dissector in the root namespace overrides
84 		 * any per-net-namespace one. When attaching to root,
85 		 * make sure we don't have any BPF program attached
86 		 * to the non-root namespaces.
87 		 */
88 		struct net *ns;
89 
90 		for_each_net(ns) {
91 			if (ns == &init_net)
92 				continue;
93 			if (rcu_access_pointer(ns->bpf.run_array[type]))
94 				return -EEXIST;
95 		}
96 	} else {
97 		/* Make sure root flow dissector is not attached
98 		 * when attaching to the non-root namespace.
99 		 */
100 		if (rcu_access_pointer(init_net.bpf.run_array[type]))
101 			return -EEXIST;
102 	}
103 
104 	return 0;
105 }
106 #endif /* CONFIG_BPF_SYSCALL */
107 
108 /**
109  * __skb_flow_get_ports - extract the upper layer ports and return them
110  * @skb: sk_buff to extract the ports from
111  * @thoff: transport header offset
112  * @ip_proto: protocol for which to get port offset
113  * @data: raw buffer pointer to the packet, if NULL use skb->data
114  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
115  *
116  * The function will try to retrieve the ports at offset thoff + poff where poff
117  * is the protocol port offset returned from proto_ports_offset
118  */
119 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
120 			    const void *data, int hlen)
121 {
122 	int poff = proto_ports_offset(ip_proto);
123 
124 	if (!data) {
125 		data = skb->data;
126 		hlen = skb_headlen(skb);
127 	}
128 
129 	if (poff >= 0) {
130 		__be32 *ports, _ports;
131 
132 		ports = __skb_header_pointer(skb, thoff + poff,
133 					     sizeof(_ports), data, hlen, &_ports);
134 		if (ports)
135 			return *ports;
136 	}
137 
138 	return 0;
139 }
140 EXPORT_SYMBOL(__skb_flow_get_ports);
141 
142 static bool icmp_has_id(u8 type)
143 {
144 	switch (type) {
145 	case ICMP_ECHO:
146 	case ICMP_ECHOREPLY:
147 	case ICMP_TIMESTAMP:
148 	case ICMP_TIMESTAMPREPLY:
149 	case ICMPV6_ECHO_REQUEST:
150 	case ICMPV6_ECHO_REPLY:
151 		return true;
152 	}
153 
154 	return false;
155 }
156 
157 /**
158  * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
159  * @skb: sk_buff to extract from
160  * @key_icmp: struct flow_dissector_key_icmp to fill
161  * @data: raw buffer pointer to the packet
162  * @thoff: offset to extract at
163  * @hlen: packet header length
164  */
165 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
166 			   struct flow_dissector_key_icmp *key_icmp,
167 			   const void *data, int thoff, int hlen)
168 {
169 	struct icmphdr *ih, _ih;
170 
171 	ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
172 	if (!ih)
173 		return;
174 
175 	key_icmp->type = ih->type;
176 	key_icmp->code = ih->code;
177 
178 	/* As we use 0 to signal that the Id field is not present,
179 	 * avoid confusion with packets without such field
180 	 */
181 	if (icmp_has_id(ih->type))
182 		key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
183 	else
184 		key_icmp->id = 0;
185 }
186 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
187 
188 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
189  * using skb_flow_get_icmp_tci().
190  */
191 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
192 				    struct flow_dissector *flow_dissector,
193 				    void *target_container, const void *data,
194 				    int thoff, int hlen)
195 {
196 	struct flow_dissector_key_icmp *key_icmp;
197 
198 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
199 		return;
200 
201 	key_icmp = skb_flow_dissector_target(flow_dissector,
202 					     FLOW_DISSECTOR_KEY_ICMP,
203 					     target_container);
204 
205 	skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
206 }
207 
208 static void __skb_flow_dissect_ah(const struct sk_buff *skb,
209 				  struct flow_dissector *flow_dissector,
210 				  void *target_container, const void *data,
211 				  int nhoff, int hlen)
212 {
213 	struct flow_dissector_key_ipsec *key_ah;
214 	struct ip_auth_hdr _hdr, *hdr;
215 
216 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
217 		return;
218 
219 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
220 	if (!hdr)
221 		return;
222 
223 	key_ah = skb_flow_dissector_target(flow_dissector,
224 					   FLOW_DISSECTOR_KEY_IPSEC,
225 					   target_container);
226 
227 	key_ah->spi = hdr->spi;
228 }
229 
230 static void __skb_flow_dissect_esp(const struct sk_buff *skb,
231 				   struct flow_dissector *flow_dissector,
232 				   void *target_container, const void *data,
233 				   int nhoff, int hlen)
234 {
235 	struct flow_dissector_key_ipsec *key_esp;
236 	struct ip_esp_hdr _hdr, *hdr;
237 
238 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
239 		return;
240 
241 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
242 	if (!hdr)
243 		return;
244 
245 	key_esp = skb_flow_dissector_target(flow_dissector,
246 					    FLOW_DISSECTOR_KEY_IPSEC,
247 					    target_container);
248 
249 	key_esp->spi = hdr->spi;
250 }
251 
252 static void __skb_flow_dissect_l2tpv3(const struct sk_buff *skb,
253 				      struct flow_dissector *flow_dissector,
254 				      void *target_container, const void *data,
255 				      int nhoff, int hlen)
256 {
257 	struct flow_dissector_key_l2tpv3 *key_l2tpv3;
258 	struct {
259 		__be32 session_id;
260 	} *hdr, _hdr;
261 
262 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3))
263 		return;
264 
265 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
266 	if (!hdr)
267 		return;
268 
269 	key_l2tpv3 = skb_flow_dissector_target(flow_dissector,
270 					       FLOW_DISSECTOR_KEY_L2TPV3,
271 					       target_container);
272 
273 	key_l2tpv3->session_id = hdr->session_id;
274 }
275 
276 void skb_flow_dissect_meta(const struct sk_buff *skb,
277 			   struct flow_dissector *flow_dissector,
278 			   void *target_container)
279 {
280 	struct flow_dissector_key_meta *meta;
281 
282 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
283 		return;
284 
285 	meta = skb_flow_dissector_target(flow_dissector,
286 					 FLOW_DISSECTOR_KEY_META,
287 					 target_container);
288 	meta->ingress_ifindex = skb->skb_iif;
289 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
290 	if (tc_skb_ext_tc_enabled()) {
291 		struct tc_skb_ext *ext;
292 
293 		ext = skb_ext_find(skb, TC_SKB_EXT);
294 		if (ext)
295 			meta->l2_miss = ext->l2_miss;
296 	}
297 #endif
298 }
299 EXPORT_SYMBOL(skb_flow_dissect_meta);
300 
301 static void
302 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
303 				   struct flow_dissector *flow_dissector,
304 				   void *target_container)
305 {
306 	struct flow_dissector_key_control *ctrl;
307 
308 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
309 		return;
310 
311 	ctrl = skb_flow_dissector_target(flow_dissector,
312 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
313 					 target_container);
314 	ctrl->addr_type = type;
315 }
316 
317 void
318 skb_flow_dissect_ct(const struct sk_buff *skb,
319 		    struct flow_dissector *flow_dissector,
320 		    void *target_container, u16 *ctinfo_map,
321 		    size_t mapsize, bool post_ct, u16 zone)
322 {
323 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
324 	struct flow_dissector_key_ct *key;
325 	enum ip_conntrack_info ctinfo;
326 	struct nf_conn_labels *cl;
327 	struct nf_conn *ct;
328 
329 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
330 		return;
331 
332 	ct = nf_ct_get(skb, &ctinfo);
333 	if (!ct && !post_ct)
334 		return;
335 
336 	key = skb_flow_dissector_target(flow_dissector,
337 					FLOW_DISSECTOR_KEY_CT,
338 					target_container);
339 
340 	if (!ct) {
341 		key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
342 				TCA_FLOWER_KEY_CT_FLAGS_INVALID;
343 		key->ct_zone = zone;
344 		return;
345 	}
346 
347 	if (ctinfo < mapsize)
348 		key->ct_state = ctinfo_map[ctinfo];
349 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
350 	key->ct_zone = ct->zone.id;
351 #endif
352 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
353 	key->ct_mark = READ_ONCE(ct->mark);
354 #endif
355 
356 	cl = nf_ct_labels_find(ct);
357 	if (cl)
358 		memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
359 #endif /* CONFIG_NF_CONNTRACK */
360 }
361 EXPORT_SYMBOL(skb_flow_dissect_ct);
362 
363 void
364 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
365 			     struct flow_dissector *flow_dissector,
366 			     void *target_container)
367 {
368 	struct ip_tunnel_info *info;
369 	struct ip_tunnel_key *key;
370 
371 	/* A quick check to see if there might be something to do. */
372 	if (!dissector_uses_key(flow_dissector,
373 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
374 	    !dissector_uses_key(flow_dissector,
375 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
376 	    !dissector_uses_key(flow_dissector,
377 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
378 	    !dissector_uses_key(flow_dissector,
379 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
380 	    !dissector_uses_key(flow_dissector,
381 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
382 	    !dissector_uses_key(flow_dissector,
383 				FLOW_DISSECTOR_KEY_ENC_IP) &&
384 	    !dissector_uses_key(flow_dissector,
385 				FLOW_DISSECTOR_KEY_ENC_OPTS) &&
386 	    !dissector_uses_key(flow_dissector,
387 				FLOW_DISSECTOR_KEY_ENC_FLAGS))
388 		return;
389 
390 	info = skb_tunnel_info(skb);
391 	if (!info)
392 		return;
393 
394 	key = &info->key;
395 
396 	switch (ip_tunnel_info_af(info)) {
397 	case AF_INET:
398 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
399 						   flow_dissector,
400 						   target_container);
401 		if (dissector_uses_key(flow_dissector,
402 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
403 			struct flow_dissector_key_ipv4_addrs *ipv4;
404 
405 			ipv4 = skb_flow_dissector_target(flow_dissector,
406 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
407 							 target_container);
408 			ipv4->src = key->u.ipv4.src;
409 			ipv4->dst = key->u.ipv4.dst;
410 		}
411 		break;
412 	case AF_INET6:
413 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
414 						   flow_dissector,
415 						   target_container);
416 		if (dissector_uses_key(flow_dissector,
417 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
418 			struct flow_dissector_key_ipv6_addrs *ipv6;
419 
420 			ipv6 = skb_flow_dissector_target(flow_dissector,
421 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
422 							 target_container);
423 			ipv6->src = key->u.ipv6.src;
424 			ipv6->dst = key->u.ipv6.dst;
425 		}
426 		break;
427 	}
428 
429 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
430 		struct flow_dissector_key_keyid *keyid;
431 
432 		keyid = skb_flow_dissector_target(flow_dissector,
433 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
434 						  target_container);
435 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
436 	}
437 
438 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
439 		struct flow_dissector_key_ports *tp;
440 
441 		tp = skb_flow_dissector_target(flow_dissector,
442 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
443 					       target_container);
444 		tp->src = key->tp_src;
445 		tp->dst = key->tp_dst;
446 	}
447 
448 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
449 		struct flow_dissector_key_ip *ip;
450 
451 		ip = skb_flow_dissector_target(flow_dissector,
452 					       FLOW_DISSECTOR_KEY_ENC_IP,
453 					       target_container);
454 		ip->tos = key->tos;
455 		ip->ttl = key->ttl;
456 	}
457 
458 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
459 		struct flow_dissector_key_enc_opts *enc_opt;
460 		IP_TUNNEL_DECLARE_FLAGS(flags) = { };
461 		u32 val;
462 
463 		enc_opt = skb_flow_dissector_target(flow_dissector,
464 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
465 						    target_container);
466 
467 		if (!info->options_len)
468 			return;
469 
470 		enc_opt->len = info->options_len;
471 		ip_tunnel_info_opts_get(enc_opt->data, info);
472 
473 		ip_tunnel_set_options_present(flags);
474 		ip_tunnel_flags_and(flags, info->key.tun_flags, flags);
475 
476 		val = find_next_bit(flags, __IP_TUNNEL_FLAG_NUM,
477 				    IP_TUNNEL_GENEVE_OPT_BIT);
478 		enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val : 0;
479 	}
480 
481 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_FLAGS)) {
482 		struct flow_dissector_key_enc_flags *enc_flags;
483 		IP_TUNNEL_DECLARE_FLAGS(flags) = {};
484 
485 		enc_flags = skb_flow_dissector_target(flow_dissector,
486 						      FLOW_DISSECTOR_KEY_ENC_FLAGS,
487 						      target_container);
488 		ip_tunnel_set_encflags_present(flags);
489 		ip_tunnel_flags_and(flags, flags, info->key.tun_flags);
490 		enc_flags->flags = bitmap_read(flags, IP_TUNNEL_CSUM_BIT, 32);
491 	}
492 }
493 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
494 
495 void skb_flow_dissect_hash(const struct sk_buff *skb,
496 			   struct flow_dissector *flow_dissector,
497 			   void *target_container)
498 {
499 	struct flow_dissector_key_hash *key;
500 
501 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
502 		return;
503 
504 	key = skb_flow_dissector_target(flow_dissector,
505 					FLOW_DISSECTOR_KEY_HASH,
506 					target_container);
507 
508 	key->hash = skb_get_hash_raw(skb);
509 }
510 EXPORT_SYMBOL(skb_flow_dissect_hash);
511 
512 static enum flow_dissect_ret
513 __skb_flow_dissect_mpls(const struct sk_buff *skb,
514 			struct flow_dissector *flow_dissector,
515 			void *target_container, const void *data, int nhoff,
516 			int hlen, int lse_index, bool *entropy_label)
517 {
518 	struct mpls_label *hdr, _hdr;
519 	u32 entry, label, bos;
520 
521 	if (!dissector_uses_key(flow_dissector,
522 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
523 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
524 		return FLOW_DISSECT_RET_OUT_GOOD;
525 
526 	if (lse_index >= FLOW_DIS_MPLS_MAX)
527 		return FLOW_DISSECT_RET_OUT_GOOD;
528 
529 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
530 				   hlen, &_hdr);
531 	if (!hdr)
532 		return FLOW_DISSECT_RET_OUT_BAD;
533 
534 	entry = ntohl(hdr->entry);
535 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
536 	bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
537 
538 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
539 		struct flow_dissector_key_mpls *key_mpls;
540 		struct flow_dissector_mpls_lse *lse;
541 
542 		key_mpls = skb_flow_dissector_target(flow_dissector,
543 						     FLOW_DISSECTOR_KEY_MPLS,
544 						     target_container);
545 		lse = &key_mpls->ls[lse_index];
546 
547 		lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
548 		lse->mpls_bos = bos;
549 		lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
550 		lse->mpls_label = label;
551 		dissector_set_mpls_lse(key_mpls, lse_index);
552 	}
553 
554 	if (*entropy_label &&
555 	    dissector_uses_key(flow_dissector,
556 			       FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
557 		struct flow_dissector_key_keyid *key_keyid;
558 
559 		key_keyid = skb_flow_dissector_target(flow_dissector,
560 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
561 						      target_container);
562 		key_keyid->keyid = cpu_to_be32(label);
563 	}
564 
565 	*entropy_label = label == MPLS_LABEL_ENTROPY;
566 
567 	return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
568 }
569 
570 static enum flow_dissect_ret
571 __skb_flow_dissect_arp(const struct sk_buff *skb,
572 		       struct flow_dissector *flow_dissector,
573 		       void *target_container, const void *data,
574 		       int nhoff, int hlen)
575 {
576 	struct flow_dissector_key_arp *key_arp;
577 	struct {
578 		unsigned char ar_sha[ETH_ALEN];
579 		unsigned char ar_sip[4];
580 		unsigned char ar_tha[ETH_ALEN];
581 		unsigned char ar_tip[4];
582 	} *arp_eth, _arp_eth;
583 	const struct arphdr *arp;
584 	struct arphdr _arp;
585 
586 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
587 		return FLOW_DISSECT_RET_OUT_GOOD;
588 
589 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
590 				   hlen, &_arp);
591 	if (!arp)
592 		return FLOW_DISSECT_RET_OUT_BAD;
593 
594 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
595 	    arp->ar_pro != htons(ETH_P_IP) ||
596 	    arp->ar_hln != ETH_ALEN ||
597 	    arp->ar_pln != 4 ||
598 	    (arp->ar_op != htons(ARPOP_REPLY) &&
599 	     arp->ar_op != htons(ARPOP_REQUEST)))
600 		return FLOW_DISSECT_RET_OUT_BAD;
601 
602 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
603 				       sizeof(_arp_eth), data,
604 				       hlen, &_arp_eth);
605 	if (!arp_eth)
606 		return FLOW_DISSECT_RET_OUT_BAD;
607 
608 	key_arp = skb_flow_dissector_target(flow_dissector,
609 					    FLOW_DISSECTOR_KEY_ARP,
610 					    target_container);
611 
612 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
613 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
614 
615 	/* Only store the lower byte of the opcode;
616 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
617 	 */
618 	key_arp->op = ntohs(arp->ar_op) & 0xff;
619 
620 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
621 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
622 
623 	return FLOW_DISSECT_RET_OUT_GOOD;
624 }
625 
626 static enum flow_dissect_ret
627 __skb_flow_dissect_cfm(const struct sk_buff *skb,
628 		       struct flow_dissector *flow_dissector,
629 		       void *target_container, const void *data,
630 		       int nhoff, int hlen)
631 {
632 	struct flow_dissector_key_cfm *key, *hdr, _hdr;
633 
634 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM))
635 		return FLOW_DISSECT_RET_OUT_GOOD;
636 
637 	hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr);
638 	if (!hdr)
639 		return FLOW_DISSECT_RET_OUT_BAD;
640 
641 	key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM,
642 					target_container);
643 
644 	key->mdl_ver = hdr->mdl_ver;
645 	key->opcode = hdr->opcode;
646 
647 	return FLOW_DISSECT_RET_OUT_GOOD;
648 }
649 
650 static enum flow_dissect_ret
651 __skb_flow_dissect_gre(const struct sk_buff *skb,
652 		       struct flow_dissector_key_control *key_control,
653 		       struct flow_dissector *flow_dissector,
654 		       void *target_container, const void *data,
655 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
656 		       unsigned int flags)
657 {
658 	struct flow_dissector_key_keyid *key_keyid;
659 	struct gre_base_hdr *hdr, _hdr;
660 	int offset = 0;
661 	u16 gre_ver;
662 
663 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
664 				   data, *p_hlen, &_hdr);
665 	if (!hdr)
666 		return FLOW_DISSECT_RET_OUT_BAD;
667 
668 	/* Only look inside GRE without routing */
669 	if (hdr->flags & GRE_ROUTING)
670 		return FLOW_DISSECT_RET_OUT_GOOD;
671 
672 	/* Only look inside GRE for version 0 and 1 */
673 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
674 	if (gre_ver > 1)
675 		return FLOW_DISSECT_RET_OUT_GOOD;
676 
677 	*p_proto = hdr->protocol;
678 	if (gre_ver) {
679 		/* Version1 must be PPTP, and check the flags */
680 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
681 			return FLOW_DISSECT_RET_OUT_GOOD;
682 	}
683 
684 	offset += sizeof(struct gre_base_hdr);
685 
686 	if (hdr->flags & GRE_CSUM)
687 		offset += sizeof_field(struct gre_full_hdr, csum) +
688 			  sizeof_field(struct gre_full_hdr, reserved1);
689 
690 	if (hdr->flags & GRE_KEY) {
691 		const __be32 *keyid;
692 		__be32 _keyid;
693 
694 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
695 					     sizeof(_keyid),
696 					     data, *p_hlen, &_keyid);
697 		if (!keyid)
698 			return FLOW_DISSECT_RET_OUT_BAD;
699 
700 		if (dissector_uses_key(flow_dissector,
701 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
702 			key_keyid = skb_flow_dissector_target(flow_dissector,
703 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
704 							      target_container);
705 			if (gre_ver == 0)
706 				key_keyid->keyid = *keyid;
707 			else
708 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
709 		}
710 		offset += sizeof_field(struct gre_full_hdr, key);
711 	}
712 
713 	if (hdr->flags & GRE_SEQ)
714 		offset += sizeof_field(struct pptp_gre_header, seq);
715 
716 	if (gre_ver == 0) {
717 		if (*p_proto == htons(ETH_P_TEB)) {
718 			const struct ethhdr *eth;
719 			struct ethhdr _eth;
720 
721 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
722 						   sizeof(_eth),
723 						   data, *p_hlen, &_eth);
724 			if (!eth)
725 				return FLOW_DISSECT_RET_OUT_BAD;
726 			*p_proto = eth->h_proto;
727 			offset += sizeof(*eth);
728 
729 			/* Cap headers that we access via pointers at the
730 			 * end of the Ethernet header as our maximum alignment
731 			 * at that point is only 2 bytes.
732 			 */
733 			if (NET_IP_ALIGN)
734 				*p_hlen = *p_nhoff + offset;
735 		}
736 	} else { /* version 1, must be PPTP */
737 		u8 _ppp_hdr[PPP_HDRLEN];
738 		u8 *ppp_hdr;
739 
740 		if (hdr->flags & GRE_ACK)
741 			offset += sizeof_field(struct pptp_gre_header, ack);
742 
743 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
744 					       sizeof(_ppp_hdr),
745 					       data, *p_hlen, _ppp_hdr);
746 		if (!ppp_hdr)
747 			return FLOW_DISSECT_RET_OUT_BAD;
748 
749 		switch (PPP_PROTOCOL(ppp_hdr)) {
750 		case PPP_IP:
751 			*p_proto = htons(ETH_P_IP);
752 			break;
753 		case PPP_IPV6:
754 			*p_proto = htons(ETH_P_IPV6);
755 			break;
756 		default:
757 			/* Could probably catch some more like MPLS */
758 			break;
759 		}
760 
761 		offset += PPP_HDRLEN;
762 	}
763 
764 	*p_nhoff += offset;
765 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
766 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
767 		return FLOW_DISSECT_RET_OUT_GOOD;
768 
769 	return FLOW_DISSECT_RET_PROTO_AGAIN;
770 }
771 
772 /**
773  * __skb_flow_dissect_batadv() - dissect batman-adv header
774  * @skb: sk_buff to with the batman-adv header
775  * @key_control: flow dissectors control key
776  * @data: raw buffer pointer to the packet, if NULL use skb->data
777  * @p_proto: pointer used to update the protocol to process next
778  * @p_nhoff: pointer used to update inner network header offset
779  * @hlen: packet header length
780  * @flags: any combination of FLOW_DISSECTOR_F_*
781  *
782  * ETH_P_BATMAN packets are tried to be dissected. Only
783  * &struct batadv_unicast packets are actually processed because they contain an
784  * inner ethernet header and are usually followed by actual network header. This
785  * allows the flow dissector to continue processing the packet.
786  *
787  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
788  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
789  *  otherwise FLOW_DISSECT_RET_OUT_BAD
790  */
791 static enum flow_dissect_ret
792 __skb_flow_dissect_batadv(const struct sk_buff *skb,
793 			  struct flow_dissector_key_control *key_control,
794 			  const void *data, __be16 *p_proto, int *p_nhoff,
795 			  int hlen, unsigned int flags)
796 {
797 	struct {
798 		struct batadv_unicast_packet batadv_unicast;
799 		struct ethhdr eth;
800 	} *hdr, _hdr;
801 
802 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
803 				   &_hdr);
804 	if (!hdr)
805 		return FLOW_DISSECT_RET_OUT_BAD;
806 
807 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
808 		return FLOW_DISSECT_RET_OUT_BAD;
809 
810 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
811 		return FLOW_DISSECT_RET_OUT_BAD;
812 
813 	*p_proto = hdr->eth.h_proto;
814 	*p_nhoff += sizeof(*hdr);
815 
816 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
817 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
818 		return FLOW_DISSECT_RET_OUT_GOOD;
819 
820 	return FLOW_DISSECT_RET_PROTO_AGAIN;
821 }
822 
823 static void
824 __skb_flow_dissect_tcp(const struct sk_buff *skb,
825 		       struct flow_dissector *flow_dissector,
826 		       void *target_container, const void *data,
827 		       int thoff, int hlen)
828 {
829 	struct flow_dissector_key_tcp *key_tcp;
830 	struct tcphdr *th, _th;
831 
832 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
833 		return;
834 
835 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
836 	if (!th)
837 		return;
838 
839 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
840 		return;
841 
842 	key_tcp = skb_flow_dissector_target(flow_dissector,
843 					    FLOW_DISSECTOR_KEY_TCP,
844 					    target_container);
845 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
846 }
847 
848 static void
849 __skb_flow_dissect_ports(const struct sk_buff *skb,
850 			 struct flow_dissector *flow_dissector,
851 			 void *target_container, const void *data,
852 			 int nhoff, u8 ip_proto, int hlen)
853 {
854 	enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
855 	struct flow_dissector_key_ports *key_ports;
856 
857 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
858 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
859 	else if (dissector_uses_key(flow_dissector,
860 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
861 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
862 
863 	if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
864 		return;
865 
866 	key_ports = skb_flow_dissector_target(flow_dissector,
867 					      dissector_ports,
868 					      target_container);
869 	key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
870 						data, hlen);
871 }
872 
873 static void
874 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
875 			struct flow_dissector *flow_dissector,
876 			void *target_container, const void *data,
877 			const struct iphdr *iph)
878 {
879 	struct flow_dissector_key_ip *key_ip;
880 
881 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
882 		return;
883 
884 	key_ip = skb_flow_dissector_target(flow_dissector,
885 					   FLOW_DISSECTOR_KEY_IP,
886 					   target_container);
887 	key_ip->tos = iph->tos;
888 	key_ip->ttl = iph->ttl;
889 }
890 
891 static void
892 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
893 			struct flow_dissector *flow_dissector,
894 			void *target_container, const void *data,
895 			const struct ipv6hdr *iph)
896 {
897 	struct flow_dissector_key_ip *key_ip;
898 
899 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
900 		return;
901 
902 	key_ip = skb_flow_dissector_target(flow_dissector,
903 					   FLOW_DISSECTOR_KEY_IP,
904 					   target_container);
905 	key_ip->tos = ipv6_get_dsfield(iph);
906 	key_ip->ttl = iph->hop_limit;
907 }
908 
909 /* Maximum number of protocol headers that can be parsed in
910  * __skb_flow_dissect
911  */
912 #define MAX_FLOW_DISSECT_HDRS	15
913 
914 static bool skb_flow_dissect_allowed(int *num_hdrs)
915 {
916 	++*num_hdrs;
917 
918 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
919 }
920 
921 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
922 				     struct flow_dissector *flow_dissector,
923 				     void *target_container)
924 {
925 	struct flow_dissector_key_ports *key_ports = NULL;
926 	struct flow_dissector_key_control *key_control;
927 	struct flow_dissector_key_basic *key_basic;
928 	struct flow_dissector_key_addrs *key_addrs;
929 	struct flow_dissector_key_tags *key_tags;
930 
931 	key_control = skb_flow_dissector_target(flow_dissector,
932 						FLOW_DISSECTOR_KEY_CONTROL,
933 						target_container);
934 	key_control->thoff = flow_keys->thoff;
935 	if (flow_keys->is_frag)
936 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
937 	if (flow_keys->is_first_frag)
938 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
939 	if (flow_keys->is_encap)
940 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
941 
942 	key_basic = skb_flow_dissector_target(flow_dissector,
943 					      FLOW_DISSECTOR_KEY_BASIC,
944 					      target_container);
945 	key_basic->n_proto = flow_keys->n_proto;
946 	key_basic->ip_proto = flow_keys->ip_proto;
947 
948 	if (flow_keys->addr_proto == ETH_P_IP &&
949 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
950 		key_addrs = skb_flow_dissector_target(flow_dissector,
951 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
952 						      target_container);
953 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
954 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
955 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
956 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
957 		   dissector_uses_key(flow_dissector,
958 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
959 		key_addrs = skb_flow_dissector_target(flow_dissector,
960 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
961 						      target_container);
962 		memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
963 		       sizeof(key_addrs->v6addrs.src));
964 		memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
965 		       sizeof(key_addrs->v6addrs.dst));
966 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
967 	}
968 
969 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
970 		key_ports = skb_flow_dissector_target(flow_dissector,
971 						      FLOW_DISSECTOR_KEY_PORTS,
972 						      target_container);
973 	else if (dissector_uses_key(flow_dissector,
974 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
975 		key_ports = skb_flow_dissector_target(flow_dissector,
976 						      FLOW_DISSECTOR_KEY_PORTS_RANGE,
977 						      target_container);
978 
979 	if (key_ports) {
980 		key_ports->src = flow_keys->sport;
981 		key_ports->dst = flow_keys->dport;
982 	}
983 
984 	if (dissector_uses_key(flow_dissector,
985 			       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
986 		key_tags = skb_flow_dissector_target(flow_dissector,
987 						     FLOW_DISSECTOR_KEY_FLOW_LABEL,
988 						     target_container);
989 		key_tags->flow_label = ntohl(flow_keys->flow_label);
990 	}
991 }
992 
993 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
994 		     __be16 proto, int nhoff, int hlen, unsigned int flags)
995 {
996 	struct bpf_flow_keys *flow_keys = ctx->flow_keys;
997 	u32 result;
998 
999 	/* Pass parameters to the BPF program */
1000 	memset(flow_keys, 0, sizeof(*flow_keys));
1001 	flow_keys->n_proto = proto;
1002 	flow_keys->nhoff = nhoff;
1003 	flow_keys->thoff = flow_keys->nhoff;
1004 
1005 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
1006 		     (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
1007 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
1008 		     (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1009 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
1010 		     (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
1011 	flow_keys->flags = flags;
1012 
1013 	result = bpf_prog_run_pin_on_cpu(prog, ctx);
1014 
1015 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1016 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1017 				   flow_keys->nhoff, hlen);
1018 
1019 	return result;
1020 }
1021 
1022 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1023 {
1024 	return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1025 }
1026 
1027 /**
1028  * __skb_flow_dissect - extract the flow_keys struct and return it
1029  * @net: associated network namespace, derived from @skb if NULL
1030  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1031  * @flow_dissector: list of keys to dissect
1032  * @target_container: target structure to put dissected values into
1033  * @data: raw buffer pointer to the packet, if NULL use skb->data
1034  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1035  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1036  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1037  * @flags: flags that control the dissection process, e.g.
1038  *         FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1039  *
1040  * The function will try to retrieve individual keys into target specified
1041  * by flow_dissector from either the skbuff or a raw buffer specified by the
1042  * rest parameters.
1043  *
1044  * Caller must take care of zeroing target container memory.
1045  */
1046 bool __skb_flow_dissect(const struct net *net,
1047 			const struct sk_buff *skb,
1048 			struct flow_dissector *flow_dissector,
1049 			void *target_container, const void *data,
1050 			__be16 proto, int nhoff, int hlen, unsigned int flags)
1051 {
1052 	struct flow_dissector_key_control *key_control;
1053 	struct flow_dissector_key_basic *key_basic;
1054 	struct flow_dissector_key_addrs *key_addrs;
1055 	struct flow_dissector_key_tags *key_tags;
1056 	struct flow_dissector_key_vlan *key_vlan;
1057 	enum flow_dissect_ret fdret;
1058 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1059 	bool mpls_el = false;
1060 	int mpls_lse = 0;
1061 	int num_hdrs = 0;
1062 	u8 ip_proto = 0;
1063 	bool ret;
1064 
1065 	if (!data) {
1066 		data = skb->data;
1067 		proto = skb_vlan_tag_present(skb) ?
1068 			 skb->vlan_proto : skb->protocol;
1069 		nhoff = skb_network_offset(skb);
1070 		hlen = skb_headlen(skb);
1071 #if IS_ENABLED(CONFIG_NET_DSA)
1072 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1073 			     proto == htons(ETH_P_XDSA))) {
1074 			struct metadata_dst *md_dst = skb_metadata_dst(skb);
1075 			const struct dsa_device_ops *ops;
1076 			int offset = 0;
1077 
1078 			ops = skb->dev->dsa_ptr->tag_ops;
1079 			/* Only DSA header taggers break flow dissection */
1080 			if (ops->needed_headroom &&
1081 			    (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1082 				if (ops->flow_dissect)
1083 					ops->flow_dissect(skb, &proto, &offset);
1084 				else
1085 					dsa_tag_generic_flow_dissect(skb,
1086 								     &proto,
1087 								     &offset);
1088 				hlen -= offset;
1089 				nhoff += offset;
1090 			}
1091 		}
1092 #endif
1093 	}
1094 
1095 	/* It is ensured by skb_flow_dissector_init() that control key will
1096 	 * be always present.
1097 	 */
1098 	key_control = skb_flow_dissector_target(flow_dissector,
1099 						FLOW_DISSECTOR_KEY_CONTROL,
1100 						target_container);
1101 
1102 	/* It is ensured by skb_flow_dissector_init() that basic key will
1103 	 * be always present.
1104 	 */
1105 	key_basic = skb_flow_dissector_target(flow_dissector,
1106 					      FLOW_DISSECTOR_KEY_BASIC,
1107 					      target_container);
1108 
1109 	if (skb) {
1110 		if (!net) {
1111 			if (skb->dev)
1112 				net = dev_net(skb->dev);
1113 			else if (skb->sk)
1114 				net = sock_net(skb->sk);
1115 		}
1116 	}
1117 
1118 	WARN_ON_ONCE(!net);
1119 	if (net) {
1120 		enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1121 		struct bpf_prog_array *run_array;
1122 
1123 		rcu_read_lock();
1124 		run_array = rcu_dereference(init_net.bpf.run_array[type]);
1125 		if (!run_array)
1126 			run_array = rcu_dereference(net->bpf.run_array[type]);
1127 
1128 		if (run_array) {
1129 			struct bpf_flow_keys flow_keys;
1130 			struct bpf_flow_dissector ctx = {
1131 				.flow_keys = &flow_keys,
1132 				.data = data,
1133 				.data_end = data + hlen,
1134 			};
1135 			__be16 n_proto = proto;
1136 			struct bpf_prog *prog;
1137 			u32 result;
1138 
1139 			if (skb) {
1140 				ctx.skb = skb;
1141 				/* we can't use 'proto' in the skb case
1142 				 * because it might be set to skb->vlan_proto
1143 				 * which has been pulled from the data
1144 				 */
1145 				n_proto = skb->protocol;
1146 			}
1147 
1148 			prog = READ_ONCE(run_array->items[0].prog);
1149 			result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1150 						  hlen, flags);
1151 			if (result == BPF_FLOW_DISSECTOR_CONTINUE)
1152 				goto dissect_continue;
1153 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1154 						 target_container);
1155 			rcu_read_unlock();
1156 			return result == BPF_OK;
1157 		}
1158 dissect_continue:
1159 		rcu_read_unlock();
1160 	}
1161 
1162 	if (dissector_uses_key(flow_dissector,
1163 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1164 		struct ethhdr *eth = eth_hdr(skb);
1165 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
1166 
1167 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1168 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
1169 							  target_container);
1170 		memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1171 	}
1172 
1173 	if (dissector_uses_key(flow_dissector,
1174 			       FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1175 		struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1176 
1177 		key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1178 							     FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1179 							     target_container);
1180 		key_num_of_vlans->num_of_vlans = 0;
1181 	}
1182 
1183 proto_again:
1184 	fdret = FLOW_DISSECT_RET_CONTINUE;
1185 
1186 	switch (proto) {
1187 	case htons(ETH_P_IP): {
1188 		const struct iphdr *iph;
1189 		struct iphdr _iph;
1190 
1191 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1192 		if (!iph || iph->ihl < 5) {
1193 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1194 			break;
1195 		}
1196 
1197 		nhoff += iph->ihl * 4;
1198 
1199 		ip_proto = iph->protocol;
1200 
1201 		if (dissector_uses_key(flow_dissector,
1202 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1203 			key_addrs = skb_flow_dissector_target(flow_dissector,
1204 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1205 							      target_container);
1206 
1207 			memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1208 			       sizeof(key_addrs->v4addrs.src));
1209 			memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1210 			       sizeof(key_addrs->v4addrs.dst));
1211 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1212 		}
1213 
1214 		__skb_flow_dissect_ipv4(skb, flow_dissector,
1215 					target_container, data, iph);
1216 
1217 		if (ip_is_fragment(iph)) {
1218 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1219 
1220 			if (iph->frag_off & htons(IP_OFFSET)) {
1221 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1222 				break;
1223 			} else {
1224 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
1225 				if (!(flags &
1226 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1227 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
1228 					break;
1229 				}
1230 			}
1231 		}
1232 
1233 		break;
1234 	}
1235 	case htons(ETH_P_IPV6): {
1236 		const struct ipv6hdr *iph;
1237 		struct ipv6hdr _iph;
1238 
1239 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1240 		if (!iph) {
1241 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1242 			break;
1243 		}
1244 
1245 		ip_proto = iph->nexthdr;
1246 		nhoff += sizeof(struct ipv6hdr);
1247 
1248 		if (dissector_uses_key(flow_dissector,
1249 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1250 			key_addrs = skb_flow_dissector_target(flow_dissector,
1251 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1252 							      target_container);
1253 
1254 			memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1255 			       sizeof(key_addrs->v6addrs.src));
1256 			memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1257 			       sizeof(key_addrs->v6addrs.dst));
1258 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1259 		}
1260 
1261 		if ((dissector_uses_key(flow_dissector,
1262 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1263 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1264 		    ip6_flowlabel(iph)) {
1265 			__be32 flow_label = ip6_flowlabel(iph);
1266 
1267 			if (dissector_uses_key(flow_dissector,
1268 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1269 				key_tags = skb_flow_dissector_target(flow_dissector,
1270 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
1271 								     target_container);
1272 				key_tags->flow_label = ntohl(flow_label);
1273 			}
1274 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1275 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1276 				break;
1277 			}
1278 		}
1279 
1280 		__skb_flow_dissect_ipv6(skb, flow_dissector,
1281 					target_container, data, iph);
1282 
1283 		break;
1284 	}
1285 	case htons(ETH_P_8021AD):
1286 	case htons(ETH_P_8021Q): {
1287 		const struct vlan_hdr *vlan = NULL;
1288 		struct vlan_hdr _vlan;
1289 		__be16 saved_vlan_tpid = proto;
1290 
1291 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1292 		    skb && skb_vlan_tag_present(skb)) {
1293 			proto = skb->protocol;
1294 		} else {
1295 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1296 						    data, hlen, &_vlan);
1297 			if (!vlan) {
1298 				fdret = FLOW_DISSECT_RET_OUT_BAD;
1299 				break;
1300 			}
1301 
1302 			proto = vlan->h_vlan_encapsulated_proto;
1303 			nhoff += sizeof(*vlan);
1304 		}
1305 
1306 		if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1307 		    !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1308 			struct flow_dissector_key_num_of_vlans *key_nvs;
1309 
1310 			key_nvs = skb_flow_dissector_target(flow_dissector,
1311 							    FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1312 							    target_container);
1313 			key_nvs->num_of_vlans++;
1314 		}
1315 
1316 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1317 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1318 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1319 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1320 		} else {
1321 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1322 			break;
1323 		}
1324 
1325 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1326 			key_vlan = skb_flow_dissector_target(flow_dissector,
1327 							     dissector_vlan,
1328 							     target_container);
1329 
1330 			if (!vlan) {
1331 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1332 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1333 			} else {
1334 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1335 					VLAN_VID_MASK;
1336 				key_vlan->vlan_priority =
1337 					(ntohs(vlan->h_vlan_TCI) &
1338 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1339 			}
1340 			key_vlan->vlan_tpid = saved_vlan_tpid;
1341 			key_vlan->vlan_eth_type = proto;
1342 		}
1343 
1344 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1345 		break;
1346 	}
1347 	case htons(ETH_P_PPP_SES): {
1348 		struct {
1349 			struct pppoe_hdr hdr;
1350 			__be16 proto;
1351 		} *hdr, _hdr;
1352 		u16 ppp_proto;
1353 
1354 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1355 		if (!hdr) {
1356 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1357 			break;
1358 		}
1359 
1360 		if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1361 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1362 			break;
1363 		}
1364 
1365 		/* least significant bit of the most significant octet
1366 		 * indicates if protocol field was compressed
1367 		 */
1368 		ppp_proto = ntohs(hdr->proto);
1369 		if (ppp_proto & 0x0100) {
1370 			ppp_proto = ppp_proto >> 8;
1371 			nhoff += PPPOE_SES_HLEN - 1;
1372 		} else {
1373 			nhoff += PPPOE_SES_HLEN;
1374 		}
1375 
1376 		if (ppp_proto == PPP_IP) {
1377 			proto = htons(ETH_P_IP);
1378 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1379 		} else if (ppp_proto == PPP_IPV6) {
1380 			proto = htons(ETH_P_IPV6);
1381 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1382 		} else if (ppp_proto == PPP_MPLS_UC) {
1383 			proto = htons(ETH_P_MPLS_UC);
1384 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1385 		} else if (ppp_proto == PPP_MPLS_MC) {
1386 			proto = htons(ETH_P_MPLS_MC);
1387 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1388 		} else if (ppp_proto_is_valid(ppp_proto)) {
1389 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1390 		} else {
1391 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1392 			break;
1393 		}
1394 
1395 		if (dissector_uses_key(flow_dissector,
1396 				       FLOW_DISSECTOR_KEY_PPPOE)) {
1397 			struct flow_dissector_key_pppoe *key_pppoe;
1398 
1399 			key_pppoe = skb_flow_dissector_target(flow_dissector,
1400 							      FLOW_DISSECTOR_KEY_PPPOE,
1401 							      target_container);
1402 			key_pppoe->session_id = hdr->hdr.sid;
1403 			key_pppoe->ppp_proto = htons(ppp_proto);
1404 			key_pppoe->type = htons(ETH_P_PPP_SES);
1405 		}
1406 		break;
1407 	}
1408 	case htons(ETH_P_TIPC): {
1409 		struct tipc_basic_hdr *hdr, _hdr;
1410 
1411 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1412 					   data, hlen, &_hdr);
1413 		if (!hdr) {
1414 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1415 			break;
1416 		}
1417 
1418 		if (dissector_uses_key(flow_dissector,
1419 				       FLOW_DISSECTOR_KEY_TIPC)) {
1420 			key_addrs = skb_flow_dissector_target(flow_dissector,
1421 							      FLOW_DISSECTOR_KEY_TIPC,
1422 							      target_container);
1423 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1424 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1425 		}
1426 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1427 		break;
1428 	}
1429 
1430 	case htons(ETH_P_MPLS_UC):
1431 	case htons(ETH_P_MPLS_MC):
1432 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1433 						target_container, data,
1434 						nhoff, hlen, mpls_lse,
1435 						&mpls_el);
1436 		nhoff += sizeof(struct mpls_label);
1437 		mpls_lse++;
1438 		break;
1439 	case htons(ETH_P_FCOE):
1440 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1441 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1442 			break;
1443 		}
1444 
1445 		nhoff += FCOE_HEADER_LEN;
1446 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1447 		break;
1448 
1449 	case htons(ETH_P_ARP):
1450 	case htons(ETH_P_RARP):
1451 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1452 					       target_container, data,
1453 					       nhoff, hlen);
1454 		break;
1455 
1456 	case htons(ETH_P_BATMAN):
1457 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1458 						  &proto, &nhoff, hlen, flags);
1459 		break;
1460 
1461 	case htons(ETH_P_1588): {
1462 		struct ptp_header *hdr, _hdr;
1463 
1464 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1465 					   hlen, &_hdr);
1466 		if (!hdr) {
1467 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1468 			break;
1469 		}
1470 
1471 		nhoff += sizeof(struct ptp_header);
1472 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1473 		break;
1474 	}
1475 
1476 	case htons(ETH_P_PRP):
1477 	case htons(ETH_P_HSR): {
1478 		struct hsr_tag *hdr, _hdr;
1479 
1480 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1481 					   &_hdr);
1482 		if (!hdr) {
1483 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1484 			break;
1485 		}
1486 
1487 		proto = hdr->encap_proto;
1488 		nhoff += HSR_HLEN;
1489 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1490 		break;
1491 	}
1492 
1493 	case htons(ETH_P_CFM):
1494 		fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1495 					       target_container, data,
1496 					       nhoff, hlen);
1497 		break;
1498 
1499 	default:
1500 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1501 		break;
1502 	}
1503 
1504 	/* Process result of proto processing */
1505 	switch (fdret) {
1506 	case FLOW_DISSECT_RET_OUT_GOOD:
1507 		goto out_good;
1508 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1509 		if (skb_flow_dissect_allowed(&num_hdrs))
1510 			goto proto_again;
1511 		goto out_good;
1512 	case FLOW_DISSECT_RET_CONTINUE:
1513 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1514 		break;
1515 	case FLOW_DISSECT_RET_OUT_BAD:
1516 	default:
1517 		goto out_bad;
1518 	}
1519 
1520 ip_proto_again:
1521 	fdret = FLOW_DISSECT_RET_CONTINUE;
1522 
1523 	switch (ip_proto) {
1524 	case IPPROTO_GRE:
1525 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1526 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1527 			break;
1528 		}
1529 
1530 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1531 					       target_container, data,
1532 					       &proto, &nhoff, &hlen, flags);
1533 		break;
1534 
1535 	case NEXTHDR_HOP:
1536 	case NEXTHDR_ROUTING:
1537 	case NEXTHDR_DEST: {
1538 		u8 _opthdr[2], *opthdr;
1539 
1540 		if (proto != htons(ETH_P_IPV6))
1541 			break;
1542 
1543 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1544 					      data, hlen, &_opthdr);
1545 		if (!opthdr) {
1546 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1547 			break;
1548 		}
1549 
1550 		ip_proto = opthdr[0];
1551 		nhoff += (opthdr[1] + 1) << 3;
1552 
1553 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1554 		break;
1555 	}
1556 	case NEXTHDR_FRAGMENT: {
1557 		struct frag_hdr _fh, *fh;
1558 
1559 		if (proto != htons(ETH_P_IPV6))
1560 			break;
1561 
1562 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1563 					  data, hlen, &_fh);
1564 
1565 		if (!fh) {
1566 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1567 			break;
1568 		}
1569 
1570 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1571 
1572 		nhoff += sizeof(_fh);
1573 		ip_proto = fh->nexthdr;
1574 
1575 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1576 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1577 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1578 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1579 				break;
1580 			}
1581 		}
1582 
1583 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1584 		break;
1585 	}
1586 	case IPPROTO_IPIP:
1587 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1588 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1589 			break;
1590 		}
1591 
1592 		proto = htons(ETH_P_IP);
1593 
1594 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1595 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1596 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1597 			break;
1598 		}
1599 
1600 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1601 		break;
1602 
1603 	case IPPROTO_IPV6:
1604 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1605 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1606 			break;
1607 		}
1608 
1609 		proto = htons(ETH_P_IPV6);
1610 
1611 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1612 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1613 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1614 			break;
1615 		}
1616 
1617 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1618 		break;
1619 
1620 
1621 	case IPPROTO_MPLS:
1622 		proto = htons(ETH_P_MPLS_UC);
1623 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1624 		break;
1625 
1626 	case IPPROTO_TCP:
1627 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1628 				       data, nhoff, hlen);
1629 		break;
1630 
1631 	case IPPROTO_ICMP:
1632 	case IPPROTO_ICMPV6:
1633 		__skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1634 					data, nhoff, hlen);
1635 		break;
1636 	case IPPROTO_L2TP:
1637 		__skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1638 					  data, nhoff, hlen);
1639 		break;
1640 	case IPPROTO_ESP:
1641 		__skb_flow_dissect_esp(skb, flow_dissector, target_container,
1642 				       data, nhoff, hlen);
1643 		break;
1644 	case IPPROTO_AH:
1645 		__skb_flow_dissect_ah(skb, flow_dissector, target_container,
1646 				      data, nhoff, hlen);
1647 		break;
1648 	default:
1649 		break;
1650 	}
1651 
1652 	if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1653 		__skb_flow_dissect_ports(skb, flow_dissector, target_container,
1654 					 data, nhoff, ip_proto, hlen);
1655 
1656 	/* Process result of IP proto processing */
1657 	switch (fdret) {
1658 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1659 		if (skb_flow_dissect_allowed(&num_hdrs))
1660 			goto proto_again;
1661 		break;
1662 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1663 		if (skb_flow_dissect_allowed(&num_hdrs))
1664 			goto ip_proto_again;
1665 		break;
1666 	case FLOW_DISSECT_RET_OUT_GOOD:
1667 	case FLOW_DISSECT_RET_CONTINUE:
1668 		break;
1669 	case FLOW_DISSECT_RET_OUT_BAD:
1670 	default:
1671 		goto out_bad;
1672 	}
1673 
1674 out_good:
1675 	ret = true;
1676 
1677 out:
1678 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1679 	key_basic->n_proto = proto;
1680 	key_basic->ip_proto = ip_proto;
1681 
1682 	return ret;
1683 
1684 out_bad:
1685 	ret = false;
1686 	goto out;
1687 }
1688 EXPORT_SYMBOL(__skb_flow_dissect);
1689 
1690 static siphash_aligned_key_t hashrnd;
1691 static __always_inline void __flow_hash_secret_init(void)
1692 {
1693 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1694 }
1695 
1696 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1697 {
1698 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1699 	return &flow->FLOW_KEYS_HASH_START_FIELD;
1700 }
1701 
1702 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1703 {
1704 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1705 
1706 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1707 
1708 	switch (flow->control.addr_type) {
1709 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1710 		diff -= sizeof(flow->addrs.v4addrs);
1711 		break;
1712 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1713 		diff -= sizeof(flow->addrs.v6addrs);
1714 		break;
1715 	case FLOW_DISSECTOR_KEY_TIPC:
1716 		diff -= sizeof(flow->addrs.tipckey);
1717 		break;
1718 	}
1719 	return sizeof(*flow) - diff;
1720 }
1721 
1722 __be32 flow_get_u32_src(const struct flow_keys *flow)
1723 {
1724 	switch (flow->control.addr_type) {
1725 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1726 		return flow->addrs.v4addrs.src;
1727 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1728 		return (__force __be32)ipv6_addr_hash(
1729 			&flow->addrs.v6addrs.src);
1730 	case FLOW_DISSECTOR_KEY_TIPC:
1731 		return flow->addrs.tipckey.key;
1732 	default:
1733 		return 0;
1734 	}
1735 }
1736 EXPORT_SYMBOL(flow_get_u32_src);
1737 
1738 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1739 {
1740 	switch (flow->control.addr_type) {
1741 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1742 		return flow->addrs.v4addrs.dst;
1743 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1744 		return (__force __be32)ipv6_addr_hash(
1745 			&flow->addrs.v6addrs.dst);
1746 	default:
1747 		return 0;
1748 	}
1749 }
1750 EXPORT_SYMBOL(flow_get_u32_dst);
1751 
1752 /* Sort the source and destination IP and the ports,
1753  * to have consistent hash within the two directions
1754  */
1755 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1756 {
1757 	int addr_diff, i;
1758 
1759 	switch (keys->control.addr_type) {
1760 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1761 		if ((__force u32)keys->addrs.v4addrs.dst <
1762 		    (__force u32)keys->addrs.v4addrs.src)
1763 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1764 
1765 		if ((__force u16)keys->ports.dst <
1766 		    (__force u16)keys->ports.src) {
1767 			swap(keys->ports.src, keys->ports.dst);
1768 		}
1769 		break;
1770 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1771 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1772 				   &keys->addrs.v6addrs.src,
1773 				   sizeof(keys->addrs.v6addrs.dst));
1774 		if (addr_diff < 0) {
1775 			for (i = 0; i < 4; i++)
1776 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1777 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1778 		}
1779 		if ((__force u16)keys->ports.dst <
1780 		    (__force u16)keys->ports.src) {
1781 			swap(keys->ports.src, keys->ports.dst);
1782 		}
1783 		break;
1784 	}
1785 }
1786 
1787 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1788 					const siphash_key_t *keyval)
1789 {
1790 	u32 hash;
1791 
1792 	__flow_hash_consistentify(keys);
1793 
1794 	hash = siphash(flow_keys_hash_start(keys),
1795 		       flow_keys_hash_length(keys), keyval);
1796 	if (!hash)
1797 		hash = 1;
1798 
1799 	return hash;
1800 }
1801 
1802 u32 flow_hash_from_keys(struct flow_keys *keys)
1803 {
1804 	__flow_hash_secret_init();
1805 	return __flow_hash_from_keys(keys, &hashrnd);
1806 }
1807 EXPORT_SYMBOL(flow_hash_from_keys);
1808 
1809 u32 flow_hash_from_keys_seed(struct flow_keys *keys,
1810 			     const siphash_key_t *keyval)
1811 {
1812 	return __flow_hash_from_keys(keys, keyval);
1813 }
1814 EXPORT_SYMBOL(flow_hash_from_keys_seed);
1815 
1816 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1817 				  struct flow_keys *keys,
1818 				  const siphash_key_t *keyval)
1819 {
1820 	skb_flow_dissect_flow_keys(skb, keys,
1821 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1822 
1823 	return __flow_hash_from_keys(keys, keyval);
1824 }
1825 
1826 struct _flow_keys_digest_data {
1827 	__be16	n_proto;
1828 	u8	ip_proto;
1829 	u8	padding;
1830 	__be32	ports;
1831 	__be32	src;
1832 	__be32	dst;
1833 };
1834 
1835 void make_flow_keys_digest(struct flow_keys_digest *digest,
1836 			   const struct flow_keys *flow)
1837 {
1838 	struct _flow_keys_digest_data *data =
1839 	    (struct _flow_keys_digest_data *)digest;
1840 
1841 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1842 
1843 	memset(digest, 0, sizeof(*digest));
1844 
1845 	data->n_proto = flow->basic.n_proto;
1846 	data->ip_proto = flow->basic.ip_proto;
1847 	data->ports = flow->ports.ports;
1848 	data->src = flow->addrs.v4addrs.src;
1849 	data->dst = flow->addrs.v4addrs.dst;
1850 }
1851 EXPORT_SYMBOL(make_flow_keys_digest);
1852 
1853 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1854 
1855 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb)
1856 {
1857 	struct flow_keys keys;
1858 
1859 	__flow_hash_secret_init();
1860 
1861 	memset(&keys, 0, sizeof(keys));
1862 	__skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric,
1863 			   &keys, NULL, 0, 0, 0, 0);
1864 
1865 	return __flow_hash_from_keys(&keys, &hashrnd);
1866 }
1867 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net);
1868 
1869 /**
1870  * __skb_get_hash_net: calculate a flow hash
1871  * @net: associated network namespace, derived from @skb if NULL
1872  * @skb: sk_buff to calculate flow hash from
1873  *
1874  * This function calculates a flow hash based on src/dst addresses
1875  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1876  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1877  * if hash is a canonical 4-tuple hash over transport ports.
1878  */
1879 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1880 {
1881 	struct flow_keys keys;
1882 	u32 hash;
1883 
1884 	memset(&keys, 0, sizeof(keys));
1885 
1886 	__skb_flow_dissect(net, skb, &flow_keys_dissector,
1887 			   &keys, NULL, 0, 0, 0,
1888 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1889 
1890 	__flow_hash_secret_init();
1891 
1892 	hash = __flow_hash_from_keys(&keys, &hashrnd);
1893 
1894 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1895 }
1896 EXPORT_SYMBOL(__skb_get_hash_net);
1897 
1898 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1899 			   const siphash_key_t *perturb)
1900 {
1901 	struct flow_keys keys;
1902 
1903 	return ___skb_get_hash(skb, &keys, perturb);
1904 }
1905 EXPORT_SYMBOL(skb_get_hash_perturb);
1906 
1907 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1908 		   const struct flow_keys_basic *keys, int hlen)
1909 {
1910 	u32 poff = keys->control.thoff;
1911 
1912 	/* skip L4 headers for fragments after the first */
1913 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1914 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1915 		return poff;
1916 
1917 	switch (keys->basic.ip_proto) {
1918 	case IPPROTO_TCP: {
1919 		/* access doff as u8 to avoid unaligned access */
1920 		const u8 *doff;
1921 		u8 _doff;
1922 
1923 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1924 					    data, hlen, &_doff);
1925 		if (!doff)
1926 			return poff;
1927 
1928 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1929 		break;
1930 	}
1931 	case IPPROTO_UDP:
1932 	case IPPROTO_UDPLITE:
1933 		poff += sizeof(struct udphdr);
1934 		break;
1935 	/* For the rest, we do not really care about header
1936 	 * extensions at this point for now.
1937 	 */
1938 	case IPPROTO_ICMP:
1939 		poff += sizeof(struct icmphdr);
1940 		break;
1941 	case IPPROTO_ICMPV6:
1942 		poff += sizeof(struct icmp6hdr);
1943 		break;
1944 	case IPPROTO_IGMP:
1945 		poff += sizeof(struct igmphdr);
1946 		break;
1947 	case IPPROTO_DCCP:
1948 		poff += sizeof(struct dccp_hdr);
1949 		break;
1950 	case IPPROTO_SCTP:
1951 		poff += sizeof(struct sctphdr);
1952 		break;
1953 	}
1954 
1955 	return poff;
1956 }
1957 
1958 /**
1959  * skb_get_poff - get the offset to the payload
1960  * @skb: sk_buff to get the payload offset from
1961  *
1962  * The function will get the offset to the payload as far as it could
1963  * be dissected.  The main user is currently BPF, so that we can dynamically
1964  * truncate packets without needing to push actual payload to the user
1965  * space and can analyze headers only, instead.
1966  */
1967 u32 skb_get_poff(const struct sk_buff *skb)
1968 {
1969 	struct flow_keys_basic keys;
1970 
1971 	if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1972 					      NULL, 0, 0, 0, 0))
1973 		return 0;
1974 
1975 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1976 }
1977 
1978 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1979 {
1980 	memset(keys, 0, sizeof(*keys));
1981 
1982 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1983 	    sizeof(keys->addrs.v6addrs.src));
1984 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1985 	    sizeof(keys->addrs.v6addrs.dst));
1986 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1987 	keys->ports.src = fl6->fl6_sport;
1988 	keys->ports.dst = fl6->fl6_dport;
1989 	keys->keyid.keyid = fl6->fl6_gre_key;
1990 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1991 	keys->basic.ip_proto = fl6->flowi6_proto;
1992 
1993 	return flow_hash_from_keys(keys);
1994 }
1995 EXPORT_SYMBOL(__get_hash_from_flowi6);
1996 
1997 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1998 	{
1999 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2000 		.offset = offsetof(struct flow_keys, control),
2001 	},
2002 	{
2003 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2004 		.offset = offsetof(struct flow_keys, basic),
2005 	},
2006 	{
2007 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2008 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
2009 	},
2010 	{
2011 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2012 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
2013 	},
2014 	{
2015 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
2016 		.offset = offsetof(struct flow_keys, addrs.tipckey),
2017 	},
2018 	{
2019 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
2020 		.offset = offsetof(struct flow_keys, ports),
2021 	},
2022 	{
2023 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
2024 		.offset = offsetof(struct flow_keys, vlan),
2025 	},
2026 	{
2027 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2028 		.offset = offsetof(struct flow_keys, tags),
2029 	},
2030 	{
2031 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2032 		.offset = offsetof(struct flow_keys, keyid),
2033 	},
2034 };
2035 
2036 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2037 	{
2038 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2039 		.offset = offsetof(struct flow_keys, control),
2040 	},
2041 	{
2042 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2043 		.offset = offsetof(struct flow_keys, basic),
2044 	},
2045 	{
2046 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2047 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
2048 	},
2049 	{
2050 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2051 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
2052 	},
2053 	{
2054 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
2055 		.offset = offsetof(struct flow_keys, ports),
2056 	},
2057 };
2058 
2059 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2060 	{
2061 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2062 		.offset = offsetof(struct flow_keys, control),
2063 	},
2064 	{
2065 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2066 		.offset = offsetof(struct flow_keys, basic),
2067 	},
2068 };
2069 
2070 struct flow_dissector flow_keys_dissector __read_mostly;
2071 EXPORT_SYMBOL(flow_keys_dissector);
2072 
2073 struct flow_dissector flow_keys_basic_dissector __read_mostly;
2074 EXPORT_SYMBOL(flow_keys_basic_dissector);
2075 
2076 static int __init init_default_flow_dissectors(void)
2077 {
2078 	skb_flow_dissector_init(&flow_keys_dissector,
2079 				flow_keys_dissector_keys,
2080 				ARRAY_SIZE(flow_keys_dissector_keys));
2081 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2082 				flow_keys_dissector_symmetric_keys,
2083 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2084 	skb_flow_dissector_init(&flow_keys_basic_dissector,
2085 				flow_keys_basic_dissector_keys,
2086 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
2087 	return 0;
2088 }
2089 core_initcall(init_default_flow_dissectors);
2090