xref: /openbmc/linux/net/core/flow_dissector.c (revision 2a9eb57e)
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 <scsi/fc/fc_fcoe.h>
31 #include <uapi/linux/batadv_packet.h>
32 #include <linux/bpf.h>
33 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
34 #include <net/netfilter/nf_conntrack_core.h>
35 #include <net/netfilter/nf_conntrack_labels.h>
36 #endif
37 #include <linux/bpf-netns.h>
38 
39 static void dissector_set_key(struct flow_dissector *flow_dissector,
40 			      enum flow_dissector_key_id key_id)
41 {
42 	flow_dissector->used_keys |= (1 << key_id);
43 }
44 
45 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
46 			     const struct flow_dissector_key *key,
47 			     unsigned int key_count)
48 {
49 	unsigned int i;
50 
51 	memset(flow_dissector, 0, sizeof(*flow_dissector));
52 
53 	for (i = 0; i < key_count; i++, key++) {
54 		/* User should make sure that every key target offset is within
55 		 * boundaries of unsigned short.
56 		 */
57 		BUG_ON(key->offset > USHRT_MAX);
58 		BUG_ON(dissector_uses_key(flow_dissector,
59 					  key->key_id));
60 
61 		dissector_set_key(flow_dissector, key->key_id);
62 		flow_dissector->offset[key->key_id] = key->offset;
63 	}
64 
65 	/* Ensure that the dissector always includes control and basic key.
66 	 * That way we are able to avoid handling lack of these in fast path.
67 	 */
68 	BUG_ON(!dissector_uses_key(flow_dissector,
69 				   FLOW_DISSECTOR_KEY_CONTROL));
70 	BUG_ON(!dissector_uses_key(flow_dissector,
71 				   FLOW_DISSECTOR_KEY_BASIC));
72 }
73 EXPORT_SYMBOL(skb_flow_dissector_init);
74 
75 #ifdef CONFIG_BPF_SYSCALL
76 int flow_dissector_bpf_prog_attach_check(struct net *net,
77 					 struct bpf_prog *prog)
78 {
79 	enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
80 
81 	if (net == &init_net) {
82 		/* BPF flow dissector in the root namespace overrides
83 		 * any per-net-namespace one. When attaching to root,
84 		 * make sure we don't have any BPF program attached
85 		 * to the non-root namespaces.
86 		 */
87 		struct net *ns;
88 
89 		for_each_net(ns) {
90 			if (ns == &init_net)
91 				continue;
92 			if (rcu_access_pointer(ns->bpf.run_array[type]))
93 				return -EEXIST;
94 		}
95 	} else {
96 		/* Make sure root flow dissector is not attached
97 		 * when attaching to the non-root namespace.
98 		 */
99 		if (rcu_access_pointer(init_net.bpf.run_array[type]))
100 			return -EEXIST;
101 	}
102 
103 	return 0;
104 }
105 #endif /* CONFIG_BPF_SYSCALL */
106 
107 /**
108  * __skb_flow_get_ports - extract the upper layer ports and return them
109  * @skb: sk_buff to extract the ports from
110  * @thoff: transport header offset
111  * @ip_proto: protocol for which to get port offset
112  * @data: raw buffer pointer to the packet, if NULL use skb->data
113  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
114  *
115  * The function will try to retrieve the ports at offset thoff + poff where poff
116  * is the protocol port offset returned from proto_ports_offset
117  */
118 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
119 			    const void *data, int hlen)
120 {
121 	int poff = proto_ports_offset(ip_proto);
122 
123 	if (!data) {
124 		data = skb->data;
125 		hlen = skb_headlen(skb);
126 	}
127 
128 	if (poff >= 0) {
129 		__be32 *ports, _ports;
130 
131 		ports = __skb_header_pointer(skb, thoff + poff,
132 					     sizeof(_ports), data, hlen, &_ports);
133 		if (ports)
134 			return *ports;
135 	}
136 
137 	return 0;
138 }
139 EXPORT_SYMBOL(__skb_flow_get_ports);
140 
141 static bool icmp_has_id(u8 type)
142 {
143 	switch (type) {
144 	case ICMP_ECHO:
145 	case ICMP_ECHOREPLY:
146 	case ICMP_TIMESTAMP:
147 	case ICMP_TIMESTAMPREPLY:
148 	case ICMPV6_ECHO_REQUEST:
149 	case ICMPV6_ECHO_REPLY:
150 		return true;
151 	}
152 
153 	return false;
154 }
155 
156 /**
157  * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
158  * @skb: sk_buff to extract from
159  * @key_icmp: struct flow_dissector_key_icmp to fill
160  * @data: raw buffer pointer to the packet
161  * @thoff: offset to extract at
162  * @hlen: packet header length
163  */
164 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
165 			   struct flow_dissector_key_icmp *key_icmp,
166 			   const void *data, int thoff, int hlen)
167 {
168 	struct icmphdr *ih, _ih;
169 
170 	ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
171 	if (!ih)
172 		return;
173 
174 	key_icmp->type = ih->type;
175 	key_icmp->code = ih->code;
176 
177 	/* As we use 0 to signal that the Id field is not present,
178 	 * avoid confusion with packets without such field
179 	 */
180 	if (icmp_has_id(ih->type))
181 		key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
182 	else
183 		key_icmp->id = 0;
184 }
185 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
186 
187 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
188  * using skb_flow_get_icmp_tci().
189  */
190 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
191 				    struct flow_dissector *flow_dissector,
192 				    void *target_container, const void *data,
193 				    int thoff, int hlen)
194 {
195 	struct flow_dissector_key_icmp *key_icmp;
196 
197 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
198 		return;
199 
200 	key_icmp = skb_flow_dissector_target(flow_dissector,
201 					     FLOW_DISSECTOR_KEY_ICMP,
202 					     target_container);
203 
204 	skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
205 }
206 
207 void skb_flow_dissect_meta(const struct sk_buff *skb,
208 			   struct flow_dissector *flow_dissector,
209 			   void *target_container)
210 {
211 	struct flow_dissector_key_meta *meta;
212 
213 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
214 		return;
215 
216 	meta = skb_flow_dissector_target(flow_dissector,
217 					 FLOW_DISSECTOR_KEY_META,
218 					 target_container);
219 	meta->ingress_ifindex = skb->skb_iif;
220 }
221 EXPORT_SYMBOL(skb_flow_dissect_meta);
222 
223 static void
224 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
225 				   struct flow_dissector *flow_dissector,
226 				   void *target_container)
227 {
228 	struct flow_dissector_key_control *ctrl;
229 
230 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
231 		return;
232 
233 	ctrl = skb_flow_dissector_target(flow_dissector,
234 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
235 					 target_container);
236 	ctrl->addr_type = type;
237 }
238 
239 void
240 skb_flow_dissect_ct(const struct sk_buff *skb,
241 		    struct flow_dissector *flow_dissector,
242 		    void *target_container, u16 *ctinfo_map,
243 		    size_t mapsize, bool post_ct, u16 zone)
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 && !post_ct)
256 		return;
257 
258 	key = skb_flow_dissector_target(flow_dissector,
259 					FLOW_DISSECTOR_KEY_CT,
260 					target_container);
261 
262 	if (!ct) {
263 		key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
264 				TCA_FLOWER_KEY_CT_FLAGS_INVALID;
265 		key->ct_zone = zone;
266 		return;
267 	}
268 
269 	if (ctinfo < mapsize)
270 		key->ct_state = ctinfo_map[ctinfo];
271 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
272 	key->ct_zone = ct->zone.id;
273 #endif
274 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
275 	key->ct_mark = ct->mark;
276 #endif
277 
278 	cl = nf_ct_labels_find(ct);
279 	if (cl)
280 		memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
281 #endif /* CONFIG_NF_CONNTRACK */
282 }
283 EXPORT_SYMBOL(skb_flow_dissect_ct);
284 
285 void
286 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
287 			     struct flow_dissector *flow_dissector,
288 			     void *target_container)
289 {
290 	struct ip_tunnel_info *info;
291 	struct ip_tunnel_key *key;
292 
293 	/* A quick check to see if there might be something to do. */
294 	if (!dissector_uses_key(flow_dissector,
295 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
296 	    !dissector_uses_key(flow_dissector,
297 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
298 	    !dissector_uses_key(flow_dissector,
299 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
300 	    !dissector_uses_key(flow_dissector,
301 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
302 	    !dissector_uses_key(flow_dissector,
303 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
304 	    !dissector_uses_key(flow_dissector,
305 				FLOW_DISSECTOR_KEY_ENC_IP) &&
306 	    !dissector_uses_key(flow_dissector,
307 				FLOW_DISSECTOR_KEY_ENC_OPTS))
308 		return;
309 
310 	info = skb_tunnel_info(skb);
311 	if (!info)
312 		return;
313 
314 	key = &info->key;
315 
316 	switch (ip_tunnel_info_af(info)) {
317 	case AF_INET:
318 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
319 						   flow_dissector,
320 						   target_container);
321 		if (dissector_uses_key(flow_dissector,
322 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
323 			struct flow_dissector_key_ipv4_addrs *ipv4;
324 
325 			ipv4 = skb_flow_dissector_target(flow_dissector,
326 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
327 							 target_container);
328 			ipv4->src = key->u.ipv4.src;
329 			ipv4->dst = key->u.ipv4.dst;
330 		}
331 		break;
332 	case AF_INET6:
333 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
334 						   flow_dissector,
335 						   target_container);
336 		if (dissector_uses_key(flow_dissector,
337 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
338 			struct flow_dissector_key_ipv6_addrs *ipv6;
339 
340 			ipv6 = skb_flow_dissector_target(flow_dissector,
341 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
342 							 target_container);
343 			ipv6->src = key->u.ipv6.src;
344 			ipv6->dst = key->u.ipv6.dst;
345 		}
346 		break;
347 	}
348 
349 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
350 		struct flow_dissector_key_keyid *keyid;
351 
352 		keyid = skb_flow_dissector_target(flow_dissector,
353 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
354 						  target_container);
355 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
356 	}
357 
358 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
359 		struct flow_dissector_key_ports *tp;
360 
361 		tp = skb_flow_dissector_target(flow_dissector,
362 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
363 					       target_container);
364 		tp->src = key->tp_src;
365 		tp->dst = key->tp_dst;
366 	}
367 
368 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
369 		struct flow_dissector_key_ip *ip;
370 
371 		ip = skb_flow_dissector_target(flow_dissector,
372 					       FLOW_DISSECTOR_KEY_ENC_IP,
373 					       target_container);
374 		ip->tos = key->tos;
375 		ip->ttl = key->ttl;
376 	}
377 
378 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
379 		struct flow_dissector_key_enc_opts *enc_opt;
380 
381 		enc_opt = skb_flow_dissector_target(flow_dissector,
382 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
383 						    target_container);
384 
385 		if (info->options_len) {
386 			enc_opt->len = info->options_len;
387 			ip_tunnel_info_opts_get(enc_opt->data, info);
388 			enc_opt->dst_opt_type = info->key.tun_flags &
389 						TUNNEL_OPTIONS_PRESENT;
390 		}
391 	}
392 }
393 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
394 
395 void skb_flow_dissect_hash(const struct sk_buff *skb,
396 			   struct flow_dissector *flow_dissector,
397 			   void *target_container)
398 {
399 	struct flow_dissector_key_hash *key;
400 
401 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
402 		return;
403 
404 	key = skb_flow_dissector_target(flow_dissector,
405 					FLOW_DISSECTOR_KEY_HASH,
406 					target_container);
407 
408 	key->hash = skb_get_hash_raw(skb);
409 }
410 EXPORT_SYMBOL(skb_flow_dissect_hash);
411 
412 static enum flow_dissect_ret
413 __skb_flow_dissect_mpls(const struct sk_buff *skb,
414 			struct flow_dissector *flow_dissector,
415 			void *target_container, const void *data, int nhoff,
416 			int hlen, int lse_index, bool *entropy_label)
417 {
418 	struct mpls_label *hdr, _hdr;
419 	u32 entry, label, bos;
420 
421 	if (!dissector_uses_key(flow_dissector,
422 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
423 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
424 		return FLOW_DISSECT_RET_OUT_GOOD;
425 
426 	if (lse_index >= FLOW_DIS_MPLS_MAX)
427 		return FLOW_DISSECT_RET_OUT_GOOD;
428 
429 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
430 				   hlen, &_hdr);
431 	if (!hdr)
432 		return FLOW_DISSECT_RET_OUT_BAD;
433 
434 	entry = ntohl(hdr->entry);
435 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
436 	bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
437 
438 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
439 		struct flow_dissector_key_mpls *key_mpls;
440 		struct flow_dissector_mpls_lse *lse;
441 
442 		key_mpls = skb_flow_dissector_target(flow_dissector,
443 						     FLOW_DISSECTOR_KEY_MPLS,
444 						     target_container);
445 		lse = &key_mpls->ls[lse_index];
446 
447 		lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
448 		lse->mpls_bos = bos;
449 		lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
450 		lse->mpls_label = label;
451 		dissector_set_mpls_lse(key_mpls, lse_index);
452 	}
453 
454 	if (*entropy_label &&
455 	    dissector_uses_key(flow_dissector,
456 			       FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
457 		struct flow_dissector_key_keyid *key_keyid;
458 
459 		key_keyid = skb_flow_dissector_target(flow_dissector,
460 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
461 						      target_container);
462 		key_keyid->keyid = cpu_to_be32(label);
463 	}
464 
465 	*entropy_label = label == MPLS_LABEL_ENTROPY;
466 
467 	return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
468 }
469 
470 static enum flow_dissect_ret
471 __skb_flow_dissect_arp(const struct sk_buff *skb,
472 		       struct flow_dissector *flow_dissector,
473 		       void *target_container, const void *data,
474 		       int nhoff, int hlen)
475 {
476 	struct flow_dissector_key_arp *key_arp;
477 	struct {
478 		unsigned char ar_sha[ETH_ALEN];
479 		unsigned char ar_sip[4];
480 		unsigned char ar_tha[ETH_ALEN];
481 		unsigned char ar_tip[4];
482 	} *arp_eth, _arp_eth;
483 	const struct arphdr *arp;
484 	struct arphdr _arp;
485 
486 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
487 		return FLOW_DISSECT_RET_OUT_GOOD;
488 
489 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
490 				   hlen, &_arp);
491 	if (!arp)
492 		return FLOW_DISSECT_RET_OUT_BAD;
493 
494 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
495 	    arp->ar_pro != htons(ETH_P_IP) ||
496 	    arp->ar_hln != ETH_ALEN ||
497 	    arp->ar_pln != 4 ||
498 	    (arp->ar_op != htons(ARPOP_REPLY) &&
499 	     arp->ar_op != htons(ARPOP_REQUEST)))
500 		return FLOW_DISSECT_RET_OUT_BAD;
501 
502 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
503 				       sizeof(_arp_eth), data,
504 				       hlen, &_arp_eth);
505 	if (!arp_eth)
506 		return FLOW_DISSECT_RET_OUT_BAD;
507 
508 	key_arp = skb_flow_dissector_target(flow_dissector,
509 					    FLOW_DISSECTOR_KEY_ARP,
510 					    target_container);
511 
512 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
513 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
514 
515 	/* Only store the lower byte of the opcode;
516 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
517 	 */
518 	key_arp->op = ntohs(arp->ar_op) & 0xff;
519 
520 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
521 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
522 
523 	return FLOW_DISSECT_RET_OUT_GOOD;
524 }
525 
526 static enum flow_dissect_ret
527 __skb_flow_dissect_gre(const struct sk_buff *skb,
528 		       struct flow_dissector_key_control *key_control,
529 		       struct flow_dissector *flow_dissector,
530 		       void *target_container, const void *data,
531 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
532 		       unsigned int flags)
533 {
534 	struct flow_dissector_key_keyid *key_keyid;
535 	struct gre_base_hdr *hdr, _hdr;
536 	int offset = 0;
537 	u16 gre_ver;
538 
539 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
540 				   data, *p_hlen, &_hdr);
541 	if (!hdr)
542 		return FLOW_DISSECT_RET_OUT_BAD;
543 
544 	/* Only look inside GRE without routing */
545 	if (hdr->flags & GRE_ROUTING)
546 		return FLOW_DISSECT_RET_OUT_GOOD;
547 
548 	/* Only look inside GRE for version 0 and 1 */
549 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
550 	if (gre_ver > 1)
551 		return FLOW_DISSECT_RET_OUT_GOOD;
552 
553 	*p_proto = hdr->protocol;
554 	if (gre_ver) {
555 		/* Version1 must be PPTP, and check the flags */
556 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
557 			return FLOW_DISSECT_RET_OUT_GOOD;
558 	}
559 
560 	offset += sizeof(struct gre_base_hdr);
561 
562 	if (hdr->flags & GRE_CSUM)
563 		offset += sizeof_field(struct gre_full_hdr, csum) +
564 			  sizeof_field(struct gre_full_hdr, reserved1);
565 
566 	if (hdr->flags & GRE_KEY) {
567 		const __be32 *keyid;
568 		__be32 _keyid;
569 
570 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
571 					     sizeof(_keyid),
572 					     data, *p_hlen, &_keyid);
573 		if (!keyid)
574 			return FLOW_DISSECT_RET_OUT_BAD;
575 
576 		if (dissector_uses_key(flow_dissector,
577 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
578 			key_keyid = skb_flow_dissector_target(flow_dissector,
579 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
580 							      target_container);
581 			if (gre_ver == 0)
582 				key_keyid->keyid = *keyid;
583 			else
584 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
585 		}
586 		offset += sizeof_field(struct gre_full_hdr, key);
587 	}
588 
589 	if (hdr->flags & GRE_SEQ)
590 		offset += sizeof_field(struct pptp_gre_header, seq);
591 
592 	if (gre_ver == 0) {
593 		if (*p_proto == htons(ETH_P_TEB)) {
594 			const struct ethhdr *eth;
595 			struct ethhdr _eth;
596 
597 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
598 						   sizeof(_eth),
599 						   data, *p_hlen, &_eth);
600 			if (!eth)
601 				return FLOW_DISSECT_RET_OUT_BAD;
602 			*p_proto = eth->h_proto;
603 			offset += sizeof(*eth);
604 
605 			/* Cap headers that we access via pointers at the
606 			 * end of the Ethernet header as our maximum alignment
607 			 * at that point is only 2 bytes.
608 			 */
609 			if (NET_IP_ALIGN)
610 				*p_hlen = *p_nhoff + offset;
611 		}
612 	} else { /* version 1, must be PPTP */
613 		u8 _ppp_hdr[PPP_HDRLEN];
614 		u8 *ppp_hdr;
615 
616 		if (hdr->flags & GRE_ACK)
617 			offset += sizeof_field(struct pptp_gre_header, ack);
618 
619 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
620 					       sizeof(_ppp_hdr),
621 					       data, *p_hlen, _ppp_hdr);
622 		if (!ppp_hdr)
623 			return FLOW_DISSECT_RET_OUT_BAD;
624 
625 		switch (PPP_PROTOCOL(ppp_hdr)) {
626 		case PPP_IP:
627 			*p_proto = htons(ETH_P_IP);
628 			break;
629 		case PPP_IPV6:
630 			*p_proto = htons(ETH_P_IPV6);
631 			break;
632 		default:
633 			/* Could probably catch some more like MPLS */
634 			break;
635 		}
636 
637 		offset += PPP_HDRLEN;
638 	}
639 
640 	*p_nhoff += offset;
641 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
642 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
643 		return FLOW_DISSECT_RET_OUT_GOOD;
644 
645 	return FLOW_DISSECT_RET_PROTO_AGAIN;
646 }
647 
648 /**
649  * __skb_flow_dissect_batadv() - dissect batman-adv header
650  * @skb: sk_buff to with the batman-adv header
651  * @key_control: flow dissectors control key
652  * @data: raw buffer pointer to the packet, if NULL use skb->data
653  * @p_proto: pointer used to update the protocol to process next
654  * @p_nhoff: pointer used to update inner network header offset
655  * @hlen: packet header length
656  * @flags: any combination of FLOW_DISSECTOR_F_*
657  *
658  * ETH_P_BATMAN packets are tried to be dissected. Only
659  * &struct batadv_unicast packets are actually processed because they contain an
660  * inner ethernet header and are usually followed by actual network header. This
661  * allows the flow dissector to continue processing the packet.
662  *
663  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
664  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
665  *  otherwise FLOW_DISSECT_RET_OUT_BAD
666  */
667 static enum flow_dissect_ret
668 __skb_flow_dissect_batadv(const struct sk_buff *skb,
669 			  struct flow_dissector_key_control *key_control,
670 			  const void *data, __be16 *p_proto, int *p_nhoff,
671 			  int hlen, unsigned int flags)
672 {
673 	struct {
674 		struct batadv_unicast_packet batadv_unicast;
675 		struct ethhdr eth;
676 	} *hdr, _hdr;
677 
678 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
679 				   &_hdr);
680 	if (!hdr)
681 		return FLOW_DISSECT_RET_OUT_BAD;
682 
683 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
684 		return FLOW_DISSECT_RET_OUT_BAD;
685 
686 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
687 		return FLOW_DISSECT_RET_OUT_BAD;
688 
689 	*p_proto = hdr->eth.h_proto;
690 	*p_nhoff += sizeof(*hdr);
691 
692 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
693 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
694 		return FLOW_DISSECT_RET_OUT_GOOD;
695 
696 	return FLOW_DISSECT_RET_PROTO_AGAIN;
697 }
698 
699 static void
700 __skb_flow_dissect_tcp(const struct sk_buff *skb,
701 		       struct flow_dissector *flow_dissector,
702 		       void *target_container, const void *data,
703 		       int thoff, int hlen)
704 {
705 	struct flow_dissector_key_tcp *key_tcp;
706 	struct tcphdr *th, _th;
707 
708 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
709 		return;
710 
711 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
712 	if (!th)
713 		return;
714 
715 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
716 		return;
717 
718 	key_tcp = skb_flow_dissector_target(flow_dissector,
719 					    FLOW_DISSECTOR_KEY_TCP,
720 					    target_container);
721 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
722 }
723 
724 static void
725 __skb_flow_dissect_ports(const struct sk_buff *skb,
726 			 struct flow_dissector *flow_dissector,
727 			 void *target_container, const void *data,
728 			 int nhoff, u8 ip_proto, int hlen)
729 {
730 	enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
731 	struct flow_dissector_key_ports *key_ports;
732 
733 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
734 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
735 	else if (dissector_uses_key(flow_dissector,
736 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
737 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
738 
739 	if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
740 		return;
741 
742 	key_ports = skb_flow_dissector_target(flow_dissector,
743 					      dissector_ports,
744 					      target_container);
745 	key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
746 						data, hlen);
747 }
748 
749 static void
750 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
751 			struct flow_dissector *flow_dissector,
752 			void *target_container, const void *data,
753 			const struct iphdr *iph)
754 {
755 	struct flow_dissector_key_ip *key_ip;
756 
757 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
758 		return;
759 
760 	key_ip = skb_flow_dissector_target(flow_dissector,
761 					   FLOW_DISSECTOR_KEY_IP,
762 					   target_container);
763 	key_ip->tos = iph->tos;
764 	key_ip->ttl = iph->ttl;
765 }
766 
767 static void
768 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
769 			struct flow_dissector *flow_dissector,
770 			void *target_container, const void *data,
771 			const struct ipv6hdr *iph)
772 {
773 	struct flow_dissector_key_ip *key_ip;
774 
775 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
776 		return;
777 
778 	key_ip = skb_flow_dissector_target(flow_dissector,
779 					   FLOW_DISSECTOR_KEY_IP,
780 					   target_container);
781 	key_ip->tos = ipv6_get_dsfield(iph);
782 	key_ip->ttl = iph->hop_limit;
783 }
784 
785 /* Maximum number of protocol headers that can be parsed in
786  * __skb_flow_dissect
787  */
788 #define MAX_FLOW_DISSECT_HDRS	15
789 
790 static bool skb_flow_dissect_allowed(int *num_hdrs)
791 {
792 	++*num_hdrs;
793 
794 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
795 }
796 
797 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
798 				     struct flow_dissector *flow_dissector,
799 				     void *target_container)
800 {
801 	struct flow_dissector_key_ports *key_ports = NULL;
802 	struct flow_dissector_key_control *key_control;
803 	struct flow_dissector_key_basic *key_basic;
804 	struct flow_dissector_key_addrs *key_addrs;
805 	struct flow_dissector_key_tags *key_tags;
806 
807 	key_control = skb_flow_dissector_target(flow_dissector,
808 						FLOW_DISSECTOR_KEY_CONTROL,
809 						target_container);
810 	key_control->thoff = flow_keys->thoff;
811 	if (flow_keys->is_frag)
812 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
813 	if (flow_keys->is_first_frag)
814 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
815 	if (flow_keys->is_encap)
816 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
817 
818 	key_basic = skb_flow_dissector_target(flow_dissector,
819 					      FLOW_DISSECTOR_KEY_BASIC,
820 					      target_container);
821 	key_basic->n_proto = flow_keys->n_proto;
822 	key_basic->ip_proto = flow_keys->ip_proto;
823 
824 	if (flow_keys->addr_proto == ETH_P_IP &&
825 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
826 		key_addrs = skb_flow_dissector_target(flow_dissector,
827 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
828 						      target_container);
829 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
830 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
831 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
832 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
833 		   dissector_uses_key(flow_dissector,
834 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
835 		key_addrs = skb_flow_dissector_target(flow_dissector,
836 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
837 						      target_container);
838 		memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
839 		       sizeof(key_addrs->v6addrs.src));
840 		memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
841 		       sizeof(key_addrs->v6addrs.dst));
842 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
843 	}
844 
845 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
846 		key_ports = skb_flow_dissector_target(flow_dissector,
847 						      FLOW_DISSECTOR_KEY_PORTS,
848 						      target_container);
849 	else if (dissector_uses_key(flow_dissector,
850 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
851 		key_ports = skb_flow_dissector_target(flow_dissector,
852 						      FLOW_DISSECTOR_KEY_PORTS_RANGE,
853 						      target_container);
854 
855 	if (key_ports) {
856 		key_ports->src = flow_keys->sport;
857 		key_ports->dst = flow_keys->dport;
858 	}
859 
860 	if (dissector_uses_key(flow_dissector,
861 			       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
862 		key_tags = skb_flow_dissector_target(flow_dissector,
863 						     FLOW_DISSECTOR_KEY_FLOW_LABEL,
864 						     target_container);
865 		key_tags->flow_label = ntohl(flow_keys->flow_label);
866 	}
867 }
868 
869 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
870 		      __be16 proto, int nhoff, int hlen, unsigned int flags)
871 {
872 	struct bpf_flow_keys *flow_keys = ctx->flow_keys;
873 	u32 result;
874 
875 	/* Pass parameters to the BPF program */
876 	memset(flow_keys, 0, sizeof(*flow_keys));
877 	flow_keys->n_proto = proto;
878 	flow_keys->nhoff = nhoff;
879 	flow_keys->thoff = flow_keys->nhoff;
880 
881 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
882 		     (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
883 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
884 		     (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
885 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
886 		     (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
887 	flow_keys->flags = flags;
888 
889 	result = bpf_prog_run_pin_on_cpu(prog, ctx);
890 
891 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
892 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
893 				   flow_keys->nhoff, hlen);
894 
895 	return result == BPF_OK;
896 }
897 
898 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
899 {
900 	return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
901 }
902 
903 /**
904  * __skb_flow_dissect - extract the flow_keys struct and return it
905  * @net: associated network namespace, derived from @skb if NULL
906  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
907  * @flow_dissector: list of keys to dissect
908  * @target_container: target structure to put dissected values into
909  * @data: raw buffer pointer to the packet, if NULL use skb->data
910  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
911  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
912  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
913  * @flags: flags that control the dissection process, e.g.
914  *         FLOW_DISSECTOR_F_STOP_AT_ENCAP.
915  *
916  * The function will try to retrieve individual keys into target specified
917  * by flow_dissector from either the skbuff or a raw buffer specified by the
918  * rest parameters.
919  *
920  * Caller must take care of zeroing target container memory.
921  */
922 bool __skb_flow_dissect(const struct net *net,
923 			const struct sk_buff *skb,
924 			struct flow_dissector *flow_dissector,
925 			void *target_container, const void *data,
926 			__be16 proto, int nhoff, int hlen, unsigned int flags)
927 {
928 	struct flow_dissector_key_control *key_control;
929 	struct flow_dissector_key_basic *key_basic;
930 	struct flow_dissector_key_addrs *key_addrs;
931 	struct flow_dissector_key_tags *key_tags;
932 	struct flow_dissector_key_vlan *key_vlan;
933 	enum flow_dissect_ret fdret;
934 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
935 	bool mpls_el = false;
936 	int mpls_lse = 0;
937 	int num_hdrs = 0;
938 	u8 ip_proto = 0;
939 	bool ret;
940 
941 	if (!data) {
942 		data = skb->data;
943 		proto = skb_vlan_tag_present(skb) ?
944 			 skb->vlan_proto : skb->protocol;
945 		nhoff = skb_network_offset(skb);
946 		hlen = skb_headlen(skb);
947 #if IS_ENABLED(CONFIG_NET_DSA)
948 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
949 			     proto == htons(ETH_P_XDSA))) {
950 			const struct dsa_device_ops *ops;
951 			int offset = 0;
952 
953 			ops = skb->dev->dsa_ptr->tag_ops;
954 			/* Only DSA header taggers break flow dissection */
955 			if (ops->needed_headroom) {
956 				if (ops->flow_dissect)
957 					ops->flow_dissect(skb, &proto, &offset);
958 				else
959 					dsa_tag_generic_flow_dissect(skb,
960 								     &proto,
961 								     &offset);
962 				hlen -= offset;
963 				nhoff += offset;
964 			}
965 		}
966 #endif
967 	}
968 
969 	/* It is ensured by skb_flow_dissector_init() that control key will
970 	 * be always present.
971 	 */
972 	key_control = skb_flow_dissector_target(flow_dissector,
973 						FLOW_DISSECTOR_KEY_CONTROL,
974 						target_container);
975 
976 	/* It is ensured by skb_flow_dissector_init() that basic key will
977 	 * be always present.
978 	 */
979 	key_basic = skb_flow_dissector_target(flow_dissector,
980 					      FLOW_DISSECTOR_KEY_BASIC,
981 					      target_container);
982 
983 	if (skb) {
984 		if (!net) {
985 			if (skb->dev)
986 				net = dev_net(skb->dev);
987 			else if (skb->sk)
988 				net = sock_net(skb->sk);
989 		}
990 	}
991 
992 	WARN_ON_ONCE(!net);
993 	if (net) {
994 		enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
995 		struct bpf_prog_array *run_array;
996 
997 		rcu_read_lock();
998 		run_array = rcu_dereference(init_net.bpf.run_array[type]);
999 		if (!run_array)
1000 			run_array = rcu_dereference(net->bpf.run_array[type]);
1001 
1002 		if (run_array) {
1003 			struct bpf_flow_keys flow_keys;
1004 			struct bpf_flow_dissector ctx = {
1005 				.flow_keys = &flow_keys,
1006 				.data = data,
1007 				.data_end = data + hlen,
1008 			};
1009 			__be16 n_proto = proto;
1010 			struct bpf_prog *prog;
1011 
1012 			if (skb) {
1013 				ctx.skb = skb;
1014 				/* we can't use 'proto' in the skb case
1015 				 * because it might be set to skb->vlan_proto
1016 				 * which has been pulled from the data
1017 				 */
1018 				n_proto = skb->protocol;
1019 			}
1020 
1021 			prog = READ_ONCE(run_array->items[0].prog);
1022 			ret = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1023 					       hlen, flags);
1024 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1025 						 target_container);
1026 			rcu_read_unlock();
1027 			return ret;
1028 		}
1029 		rcu_read_unlock();
1030 	}
1031 
1032 	if (dissector_uses_key(flow_dissector,
1033 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1034 		struct ethhdr *eth = eth_hdr(skb);
1035 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
1036 
1037 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1038 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
1039 							  target_container);
1040 		memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1041 	}
1042 
1043 	if (dissector_uses_key(flow_dissector,
1044 			       FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1045 		struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1046 
1047 		key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1048 							     FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1049 							     target_container);
1050 		key_num_of_vlans->num_of_vlans = 0;
1051 	}
1052 
1053 proto_again:
1054 	fdret = FLOW_DISSECT_RET_CONTINUE;
1055 
1056 	switch (proto) {
1057 	case htons(ETH_P_IP): {
1058 		const struct iphdr *iph;
1059 		struct iphdr _iph;
1060 
1061 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1062 		if (!iph || iph->ihl < 5) {
1063 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1064 			break;
1065 		}
1066 
1067 		nhoff += iph->ihl * 4;
1068 
1069 		ip_proto = iph->protocol;
1070 
1071 		if (dissector_uses_key(flow_dissector,
1072 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1073 			key_addrs = skb_flow_dissector_target(flow_dissector,
1074 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1075 							      target_container);
1076 
1077 			memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1078 			       sizeof(key_addrs->v4addrs.src));
1079 			memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1080 			       sizeof(key_addrs->v4addrs.dst));
1081 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1082 		}
1083 
1084 		__skb_flow_dissect_ipv4(skb, flow_dissector,
1085 					target_container, data, iph);
1086 
1087 		if (ip_is_fragment(iph)) {
1088 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1089 
1090 			if (iph->frag_off & htons(IP_OFFSET)) {
1091 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1092 				break;
1093 			} else {
1094 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
1095 				if (!(flags &
1096 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1097 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
1098 					break;
1099 				}
1100 			}
1101 		}
1102 
1103 		break;
1104 	}
1105 	case htons(ETH_P_IPV6): {
1106 		const struct ipv6hdr *iph;
1107 		struct ipv6hdr _iph;
1108 
1109 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1110 		if (!iph) {
1111 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1112 			break;
1113 		}
1114 
1115 		ip_proto = iph->nexthdr;
1116 		nhoff += sizeof(struct ipv6hdr);
1117 
1118 		if (dissector_uses_key(flow_dissector,
1119 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1120 			key_addrs = skb_flow_dissector_target(flow_dissector,
1121 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1122 							      target_container);
1123 
1124 			memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1125 			       sizeof(key_addrs->v6addrs.src));
1126 			memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1127 			       sizeof(key_addrs->v6addrs.dst));
1128 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1129 		}
1130 
1131 		if ((dissector_uses_key(flow_dissector,
1132 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1133 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1134 		    ip6_flowlabel(iph)) {
1135 			__be32 flow_label = ip6_flowlabel(iph);
1136 
1137 			if (dissector_uses_key(flow_dissector,
1138 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1139 				key_tags = skb_flow_dissector_target(flow_dissector,
1140 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
1141 								     target_container);
1142 				key_tags->flow_label = ntohl(flow_label);
1143 			}
1144 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1145 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1146 				break;
1147 			}
1148 		}
1149 
1150 		__skb_flow_dissect_ipv6(skb, flow_dissector,
1151 					target_container, data, iph);
1152 
1153 		break;
1154 	}
1155 	case htons(ETH_P_8021AD):
1156 	case htons(ETH_P_8021Q): {
1157 		const struct vlan_hdr *vlan = NULL;
1158 		struct vlan_hdr _vlan;
1159 		__be16 saved_vlan_tpid = proto;
1160 
1161 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1162 		    skb && skb_vlan_tag_present(skb)) {
1163 			proto = skb->protocol;
1164 		} else {
1165 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1166 						    data, hlen, &_vlan);
1167 			if (!vlan) {
1168 				fdret = FLOW_DISSECT_RET_OUT_BAD;
1169 				break;
1170 			}
1171 
1172 			proto = vlan->h_vlan_encapsulated_proto;
1173 			nhoff += sizeof(*vlan);
1174 		}
1175 
1176 		if (dissector_uses_key(flow_dissector,
1177 				       FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1178 			struct flow_dissector_key_num_of_vlans *key_nvs;
1179 
1180 			key_nvs = skb_flow_dissector_target(flow_dissector,
1181 							    FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1182 							    target_container);
1183 			key_nvs->num_of_vlans++;
1184 		}
1185 
1186 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1187 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1188 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1189 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1190 		} else {
1191 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1192 			break;
1193 		}
1194 
1195 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1196 			key_vlan = skb_flow_dissector_target(flow_dissector,
1197 							     dissector_vlan,
1198 							     target_container);
1199 
1200 			if (!vlan) {
1201 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1202 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1203 			} else {
1204 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1205 					VLAN_VID_MASK;
1206 				key_vlan->vlan_priority =
1207 					(ntohs(vlan->h_vlan_TCI) &
1208 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1209 			}
1210 			key_vlan->vlan_tpid = saved_vlan_tpid;
1211 			key_vlan->vlan_eth_type = proto;
1212 		}
1213 
1214 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1215 		break;
1216 	}
1217 	case htons(ETH_P_PPP_SES): {
1218 		struct {
1219 			struct pppoe_hdr hdr;
1220 			__be16 proto;
1221 		} *hdr, _hdr;
1222 		u16 ppp_proto;
1223 
1224 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1225 		if (!hdr) {
1226 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1227 			break;
1228 		}
1229 
1230 		if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1231 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1232 			break;
1233 		}
1234 
1235 		/* least significant bit of the most significant octet
1236 		 * indicates if protocol field was compressed
1237 		 */
1238 		ppp_proto = ntohs(hdr->proto);
1239 		if (ppp_proto & 0x0100) {
1240 			ppp_proto = ppp_proto >> 8;
1241 			nhoff += PPPOE_SES_HLEN - 1;
1242 		} else {
1243 			nhoff += PPPOE_SES_HLEN;
1244 		}
1245 
1246 		if (ppp_proto == PPP_IP) {
1247 			proto = htons(ETH_P_IP);
1248 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1249 		} else if (ppp_proto == PPP_IPV6) {
1250 			proto = htons(ETH_P_IPV6);
1251 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1252 		} else if (ppp_proto == PPP_MPLS_UC) {
1253 			proto = htons(ETH_P_MPLS_UC);
1254 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1255 		} else if (ppp_proto == PPP_MPLS_MC) {
1256 			proto = htons(ETH_P_MPLS_MC);
1257 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1258 		} else if (ppp_proto_is_valid(ppp_proto)) {
1259 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1260 		} else {
1261 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1262 			break;
1263 		}
1264 
1265 		if (dissector_uses_key(flow_dissector,
1266 				       FLOW_DISSECTOR_KEY_PPPOE)) {
1267 			struct flow_dissector_key_pppoe *key_pppoe;
1268 
1269 			key_pppoe = skb_flow_dissector_target(flow_dissector,
1270 							      FLOW_DISSECTOR_KEY_PPPOE,
1271 							      target_container);
1272 			key_pppoe->session_id = hdr->hdr.sid;
1273 			key_pppoe->ppp_proto = htons(ppp_proto);
1274 			key_pppoe->type = htons(ETH_P_PPP_SES);
1275 		}
1276 		break;
1277 	}
1278 	case htons(ETH_P_TIPC): {
1279 		struct tipc_basic_hdr *hdr, _hdr;
1280 
1281 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1282 					   data, hlen, &_hdr);
1283 		if (!hdr) {
1284 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1285 			break;
1286 		}
1287 
1288 		if (dissector_uses_key(flow_dissector,
1289 				       FLOW_DISSECTOR_KEY_TIPC)) {
1290 			key_addrs = skb_flow_dissector_target(flow_dissector,
1291 							      FLOW_DISSECTOR_KEY_TIPC,
1292 							      target_container);
1293 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1294 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1295 		}
1296 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1297 		break;
1298 	}
1299 
1300 	case htons(ETH_P_MPLS_UC):
1301 	case htons(ETH_P_MPLS_MC):
1302 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1303 						target_container, data,
1304 						nhoff, hlen, mpls_lse,
1305 						&mpls_el);
1306 		nhoff += sizeof(struct mpls_label);
1307 		mpls_lse++;
1308 		break;
1309 	case htons(ETH_P_FCOE):
1310 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1311 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1312 			break;
1313 		}
1314 
1315 		nhoff += FCOE_HEADER_LEN;
1316 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1317 		break;
1318 
1319 	case htons(ETH_P_ARP):
1320 	case htons(ETH_P_RARP):
1321 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1322 					       target_container, data,
1323 					       nhoff, hlen);
1324 		break;
1325 
1326 	case htons(ETH_P_BATMAN):
1327 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1328 						  &proto, &nhoff, hlen, flags);
1329 		break;
1330 
1331 	case htons(ETH_P_1588): {
1332 		struct ptp_header *hdr, _hdr;
1333 
1334 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1335 					   hlen, &_hdr);
1336 		if (!hdr) {
1337 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1338 			break;
1339 		}
1340 
1341 		nhoff += ntohs(hdr->message_length);
1342 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1343 		break;
1344 	}
1345 
1346 	case htons(ETH_P_PRP):
1347 	case htons(ETH_P_HSR): {
1348 		struct hsr_tag *hdr, _hdr;
1349 
1350 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1351 					   &_hdr);
1352 		if (!hdr) {
1353 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1354 			break;
1355 		}
1356 
1357 		proto = hdr->encap_proto;
1358 		nhoff += HSR_HLEN;
1359 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1360 		break;
1361 	}
1362 
1363 	default:
1364 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1365 		break;
1366 	}
1367 
1368 	/* Process result of proto processing */
1369 	switch (fdret) {
1370 	case FLOW_DISSECT_RET_OUT_GOOD:
1371 		goto out_good;
1372 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1373 		if (skb_flow_dissect_allowed(&num_hdrs))
1374 			goto proto_again;
1375 		goto out_good;
1376 	case FLOW_DISSECT_RET_CONTINUE:
1377 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1378 		break;
1379 	case FLOW_DISSECT_RET_OUT_BAD:
1380 	default:
1381 		goto out_bad;
1382 	}
1383 
1384 ip_proto_again:
1385 	fdret = FLOW_DISSECT_RET_CONTINUE;
1386 
1387 	switch (ip_proto) {
1388 	case IPPROTO_GRE:
1389 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1390 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1391 			break;
1392 		}
1393 
1394 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1395 					       target_container, data,
1396 					       &proto, &nhoff, &hlen, flags);
1397 		break;
1398 
1399 	case NEXTHDR_HOP:
1400 	case NEXTHDR_ROUTING:
1401 	case NEXTHDR_DEST: {
1402 		u8 _opthdr[2], *opthdr;
1403 
1404 		if (proto != htons(ETH_P_IPV6))
1405 			break;
1406 
1407 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1408 					      data, hlen, &_opthdr);
1409 		if (!opthdr) {
1410 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1411 			break;
1412 		}
1413 
1414 		ip_proto = opthdr[0];
1415 		nhoff += (opthdr[1] + 1) << 3;
1416 
1417 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1418 		break;
1419 	}
1420 	case NEXTHDR_FRAGMENT: {
1421 		struct frag_hdr _fh, *fh;
1422 
1423 		if (proto != htons(ETH_P_IPV6))
1424 			break;
1425 
1426 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1427 					  data, hlen, &_fh);
1428 
1429 		if (!fh) {
1430 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1431 			break;
1432 		}
1433 
1434 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1435 
1436 		nhoff += sizeof(_fh);
1437 		ip_proto = fh->nexthdr;
1438 
1439 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1440 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1441 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1442 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1443 				break;
1444 			}
1445 		}
1446 
1447 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1448 		break;
1449 	}
1450 	case IPPROTO_IPIP:
1451 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1452 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1453 			break;
1454 		}
1455 
1456 		proto = htons(ETH_P_IP);
1457 
1458 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1459 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1460 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1461 			break;
1462 		}
1463 
1464 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1465 		break;
1466 
1467 	case IPPROTO_IPV6:
1468 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1469 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1470 			break;
1471 		}
1472 
1473 		proto = htons(ETH_P_IPV6);
1474 
1475 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1476 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1477 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1478 			break;
1479 		}
1480 
1481 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1482 		break;
1483 
1484 
1485 	case IPPROTO_MPLS:
1486 		proto = htons(ETH_P_MPLS_UC);
1487 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1488 		break;
1489 
1490 	case IPPROTO_TCP:
1491 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1492 				       data, nhoff, hlen);
1493 		break;
1494 
1495 	case IPPROTO_ICMP:
1496 	case IPPROTO_ICMPV6:
1497 		__skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1498 					data, nhoff, hlen);
1499 		break;
1500 
1501 	default:
1502 		break;
1503 	}
1504 
1505 	if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1506 		__skb_flow_dissect_ports(skb, flow_dissector, target_container,
1507 					 data, nhoff, ip_proto, hlen);
1508 
1509 	/* Process result of IP proto processing */
1510 	switch (fdret) {
1511 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1512 		if (skb_flow_dissect_allowed(&num_hdrs))
1513 			goto proto_again;
1514 		break;
1515 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1516 		if (skb_flow_dissect_allowed(&num_hdrs))
1517 			goto ip_proto_again;
1518 		break;
1519 	case FLOW_DISSECT_RET_OUT_GOOD:
1520 	case FLOW_DISSECT_RET_CONTINUE:
1521 		break;
1522 	case FLOW_DISSECT_RET_OUT_BAD:
1523 	default:
1524 		goto out_bad;
1525 	}
1526 
1527 out_good:
1528 	ret = true;
1529 
1530 out:
1531 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1532 	key_basic->n_proto = proto;
1533 	key_basic->ip_proto = ip_proto;
1534 
1535 	return ret;
1536 
1537 out_bad:
1538 	ret = false;
1539 	goto out;
1540 }
1541 EXPORT_SYMBOL(__skb_flow_dissect);
1542 
1543 static siphash_aligned_key_t hashrnd;
1544 static __always_inline void __flow_hash_secret_init(void)
1545 {
1546 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1547 }
1548 
1549 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1550 {
1551 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1552 	return &flow->FLOW_KEYS_HASH_START_FIELD;
1553 }
1554 
1555 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1556 {
1557 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1558 
1559 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1560 
1561 	switch (flow->control.addr_type) {
1562 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1563 		diff -= sizeof(flow->addrs.v4addrs);
1564 		break;
1565 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1566 		diff -= sizeof(flow->addrs.v6addrs);
1567 		break;
1568 	case FLOW_DISSECTOR_KEY_TIPC:
1569 		diff -= sizeof(flow->addrs.tipckey);
1570 		break;
1571 	}
1572 	return sizeof(*flow) - diff;
1573 }
1574 
1575 __be32 flow_get_u32_src(const struct flow_keys *flow)
1576 {
1577 	switch (flow->control.addr_type) {
1578 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1579 		return flow->addrs.v4addrs.src;
1580 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1581 		return (__force __be32)ipv6_addr_hash(
1582 			&flow->addrs.v6addrs.src);
1583 	case FLOW_DISSECTOR_KEY_TIPC:
1584 		return flow->addrs.tipckey.key;
1585 	default:
1586 		return 0;
1587 	}
1588 }
1589 EXPORT_SYMBOL(flow_get_u32_src);
1590 
1591 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1592 {
1593 	switch (flow->control.addr_type) {
1594 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1595 		return flow->addrs.v4addrs.dst;
1596 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1597 		return (__force __be32)ipv6_addr_hash(
1598 			&flow->addrs.v6addrs.dst);
1599 	default:
1600 		return 0;
1601 	}
1602 }
1603 EXPORT_SYMBOL(flow_get_u32_dst);
1604 
1605 /* Sort the source and destination IP and the ports,
1606  * to have consistent hash within the two directions
1607  */
1608 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1609 {
1610 	int addr_diff, i;
1611 
1612 	switch (keys->control.addr_type) {
1613 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1614 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1615 			    (__force u32)keys->addrs.v4addrs.src;
1616 		if (addr_diff < 0)
1617 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1618 
1619 		if ((__force u16)keys->ports.dst <
1620 		    (__force u16)keys->ports.src) {
1621 			swap(keys->ports.src, keys->ports.dst);
1622 		}
1623 		break;
1624 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1625 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1626 				   &keys->addrs.v6addrs.src,
1627 				   sizeof(keys->addrs.v6addrs.dst));
1628 		if (addr_diff < 0) {
1629 			for (i = 0; i < 4; i++)
1630 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1631 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1632 		}
1633 		if ((__force u16)keys->ports.dst <
1634 		    (__force u16)keys->ports.src) {
1635 			swap(keys->ports.src, keys->ports.dst);
1636 		}
1637 		break;
1638 	}
1639 }
1640 
1641 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1642 					const siphash_key_t *keyval)
1643 {
1644 	u32 hash;
1645 
1646 	__flow_hash_consistentify(keys);
1647 
1648 	hash = siphash(flow_keys_hash_start(keys),
1649 		       flow_keys_hash_length(keys), keyval);
1650 	if (!hash)
1651 		hash = 1;
1652 
1653 	return hash;
1654 }
1655 
1656 u32 flow_hash_from_keys(struct flow_keys *keys)
1657 {
1658 	__flow_hash_secret_init();
1659 	return __flow_hash_from_keys(keys, &hashrnd);
1660 }
1661 EXPORT_SYMBOL(flow_hash_from_keys);
1662 
1663 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1664 				  struct flow_keys *keys,
1665 				  const siphash_key_t *keyval)
1666 {
1667 	skb_flow_dissect_flow_keys(skb, keys,
1668 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1669 
1670 	return __flow_hash_from_keys(keys, keyval);
1671 }
1672 
1673 struct _flow_keys_digest_data {
1674 	__be16	n_proto;
1675 	u8	ip_proto;
1676 	u8	padding;
1677 	__be32	ports;
1678 	__be32	src;
1679 	__be32	dst;
1680 };
1681 
1682 void make_flow_keys_digest(struct flow_keys_digest *digest,
1683 			   const struct flow_keys *flow)
1684 {
1685 	struct _flow_keys_digest_data *data =
1686 	    (struct _flow_keys_digest_data *)digest;
1687 
1688 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1689 
1690 	memset(digest, 0, sizeof(*digest));
1691 
1692 	data->n_proto = flow->basic.n_proto;
1693 	data->ip_proto = flow->basic.ip_proto;
1694 	data->ports = flow->ports.ports;
1695 	data->src = flow->addrs.v4addrs.src;
1696 	data->dst = flow->addrs.v4addrs.dst;
1697 }
1698 EXPORT_SYMBOL(make_flow_keys_digest);
1699 
1700 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1701 
1702 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1703 {
1704 	struct flow_keys keys;
1705 
1706 	__flow_hash_secret_init();
1707 
1708 	memset(&keys, 0, sizeof(keys));
1709 	__skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1710 			   &keys, NULL, 0, 0, 0,
1711 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1712 
1713 	return __flow_hash_from_keys(&keys, &hashrnd);
1714 }
1715 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1716 
1717 /**
1718  * __skb_get_hash: calculate a flow hash
1719  * @skb: sk_buff to calculate flow hash from
1720  *
1721  * This function calculates a flow hash based on src/dst addresses
1722  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1723  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1724  * if hash is a canonical 4-tuple hash over transport ports.
1725  */
1726 void __skb_get_hash(struct sk_buff *skb)
1727 {
1728 	struct flow_keys keys;
1729 	u32 hash;
1730 
1731 	__flow_hash_secret_init();
1732 
1733 	hash = ___skb_get_hash(skb, &keys, &hashrnd);
1734 
1735 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1736 }
1737 EXPORT_SYMBOL(__skb_get_hash);
1738 
1739 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1740 			   const siphash_key_t *perturb)
1741 {
1742 	struct flow_keys keys;
1743 
1744 	return ___skb_get_hash(skb, &keys, perturb);
1745 }
1746 EXPORT_SYMBOL(skb_get_hash_perturb);
1747 
1748 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1749 		   const struct flow_keys_basic *keys, int hlen)
1750 {
1751 	u32 poff = keys->control.thoff;
1752 
1753 	/* skip L4 headers for fragments after the first */
1754 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1755 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1756 		return poff;
1757 
1758 	switch (keys->basic.ip_proto) {
1759 	case IPPROTO_TCP: {
1760 		/* access doff as u8 to avoid unaligned access */
1761 		const u8 *doff;
1762 		u8 _doff;
1763 
1764 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1765 					    data, hlen, &_doff);
1766 		if (!doff)
1767 			return poff;
1768 
1769 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1770 		break;
1771 	}
1772 	case IPPROTO_UDP:
1773 	case IPPROTO_UDPLITE:
1774 		poff += sizeof(struct udphdr);
1775 		break;
1776 	/* For the rest, we do not really care about header
1777 	 * extensions at this point for now.
1778 	 */
1779 	case IPPROTO_ICMP:
1780 		poff += sizeof(struct icmphdr);
1781 		break;
1782 	case IPPROTO_ICMPV6:
1783 		poff += sizeof(struct icmp6hdr);
1784 		break;
1785 	case IPPROTO_IGMP:
1786 		poff += sizeof(struct igmphdr);
1787 		break;
1788 	case IPPROTO_DCCP:
1789 		poff += sizeof(struct dccp_hdr);
1790 		break;
1791 	case IPPROTO_SCTP:
1792 		poff += sizeof(struct sctphdr);
1793 		break;
1794 	}
1795 
1796 	return poff;
1797 }
1798 
1799 /**
1800  * skb_get_poff - get the offset to the payload
1801  * @skb: sk_buff to get the payload offset from
1802  *
1803  * The function will get the offset to the payload as far as it could
1804  * be dissected.  The main user is currently BPF, so that we can dynamically
1805  * truncate packets without needing to push actual payload to the user
1806  * space and can analyze headers only, instead.
1807  */
1808 u32 skb_get_poff(const struct sk_buff *skb)
1809 {
1810 	struct flow_keys_basic keys;
1811 
1812 	if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1813 					      NULL, 0, 0, 0, 0))
1814 		return 0;
1815 
1816 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1817 }
1818 
1819 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1820 {
1821 	memset(keys, 0, sizeof(*keys));
1822 
1823 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1824 	    sizeof(keys->addrs.v6addrs.src));
1825 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1826 	    sizeof(keys->addrs.v6addrs.dst));
1827 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1828 	keys->ports.src = fl6->fl6_sport;
1829 	keys->ports.dst = fl6->fl6_dport;
1830 	keys->keyid.keyid = fl6->fl6_gre_key;
1831 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1832 	keys->basic.ip_proto = fl6->flowi6_proto;
1833 
1834 	return flow_hash_from_keys(keys);
1835 }
1836 EXPORT_SYMBOL(__get_hash_from_flowi6);
1837 
1838 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1839 	{
1840 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1841 		.offset = offsetof(struct flow_keys, control),
1842 	},
1843 	{
1844 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1845 		.offset = offsetof(struct flow_keys, basic),
1846 	},
1847 	{
1848 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1849 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1850 	},
1851 	{
1852 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1853 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1854 	},
1855 	{
1856 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1857 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1858 	},
1859 	{
1860 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1861 		.offset = offsetof(struct flow_keys, ports),
1862 	},
1863 	{
1864 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1865 		.offset = offsetof(struct flow_keys, vlan),
1866 	},
1867 	{
1868 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1869 		.offset = offsetof(struct flow_keys, tags),
1870 	},
1871 	{
1872 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1873 		.offset = offsetof(struct flow_keys, keyid),
1874 	},
1875 };
1876 
1877 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1878 	{
1879 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1880 		.offset = offsetof(struct flow_keys, control),
1881 	},
1882 	{
1883 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1884 		.offset = offsetof(struct flow_keys, basic),
1885 	},
1886 	{
1887 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1888 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1889 	},
1890 	{
1891 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1892 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1893 	},
1894 	{
1895 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1896 		.offset = offsetof(struct flow_keys, ports),
1897 	},
1898 };
1899 
1900 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1901 	{
1902 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1903 		.offset = offsetof(struct flow_keys, control),
1904 	},
1905 	{
1906 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1907 		.offset = offsetof(struct flow_keys, basic),
1908 	},
1909 };
1910 
1911 struct flow_dissector flow_keys_dissector __read_mostly;
1912 EXPORT_SYMBOL(flow_keys_dissector);
1913 
1914 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1915 EXPORT_SYMBOL(flow_keys_basic_dissector);
1916 
1917 static int __init init_default_flow_dissectors(void)
1918 {
1919 	skb_flow_dissector_init(&flow_keys_dissector,
1920 				flow_keys_dissector_keys,
1921 				ARRAY_SIZE(flow_keys_dissector_keys));
1922 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1923 				flow_keys_dissector_symmetric_keys,
1924 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1925 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1926 				flow_keys_basic_dissector_keys,
1927 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1928 	return 0;
1929 }
1930 core_initcall(init_default_flow_dissectors);
1931