xref: /openbmc/linux/net/core/flow_dissector.c (revision efe4a1ac)
1 #include <linux/kernel.h>
2 #include <linux/skbuff.h>
3 #include <linux/export.h>
4 #include <linux/ip.h>
5 #include <linux/ipv6.h>
6 #include <linux/if_vlan.h>
7 #include <net/ip.h>
8 #include <net/ipv6.h>
9 #include <net/gre.h>
10 #include <net/pptp.h>
11 #include <linux/igmp.h>
12 #include <linux/icmp.h>
13 #include <linux/sctp.h>
14 #include <linux/dccp.h>
15 #include <linux/if_tunnel.h>
16 #include <linux/if_pppox.h>
17 #include <linux/ppp_defs.h>
18 #include <linux/stddef.h>
19 #include <linux/if_ether.h>
20 #include <linux/mpls.h>
21 #include <net/flow_dissector.h>
22 #include <scsi/fc/fc_fcoe.h>
23 
24 static void dissector_set_key(struct flow_dissector *flow_dissector,
25 			      enum flow_dissector_key_id key_id)
26 {
27 	flow_dissector->used_keys |= (1 << key_id);
28 }
29 
30 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
31 			     const struct flow_dissector_key *key,
32 			     unsigned int key_count)
33 {
34 	unsigned int i;
35 
36 	memset(flow_dissector, 0, sizeof(*flow_dissector));
37 
38 	for (i = 0; i < key_count; i++, key++) {
39 		/* User should make sure that every key target offset is withing
40 		 * boundaries of unsigned short.
41 		 */
42 		BUG_ON(key->offset > USHRT_MAX);
43 		BUG_ON(dissector_uses_key(flow_dissector,
44 					  key->key_id));
45 
46 		dissector_set_key(flow_dissector, key->key_id);
47 		flow_dissector->offset[key->key_id] = key->offset;
48 	}
49 
50 	/* Ensure that the dissector always includes control and basic key.
51 	 * That way we are able to avoid handling lack of these in fast path.
52 	 */
53 	BUG_ON(!dissector_uses_key(flow_dissector,
54 				   FLOW_DISSECTOR_KEY_CONTROL));
55 	BUG_ON(!dissector_uses_key(flow_dissector,
56 				   FLOW_DISSECTOR_KEY_BASIC));
57 }
58 EXPORT_SYMBOL(skb_flow_dissector_init);
59 
60 /**
61  * skb_flow_get_be16 - extract be16 entity
62  * @skb: sk_buff to extract from
63  * @poff: offset to extract at
64  * @data: raw buffer pointer to the packet
65  * @hlen: packet header length
66  *
67  * The function will try to retrieve a be32 entity at
68  * offset poff
69  */
70 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
71 				void *data, int hlen)
72 {
73 	__be16 *u, _u;
74 
75 	u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
76 	if (u)
77 		return *u;
78 
79 	return 0;
80 }
81 
82 /**
83  * __skb_flow_get_ports - extract the upper layer ports and return them
84  * @skb: sk_buff to extract the ports from
85  * @thoff: transport header offset
86  * @ip_proto: protocol for which to get port offset
87  * @data: raw buffer pointer to the packet, if NULL use skb->data
88  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
89  *
90  * The function will try to retrieve the ports at offset thoff + poff where poff
91  * is the protocol port offset returned from proto_ports_offset
92  */
93 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
94 			    void *data, int hlen)
95 {
96 	int poff = proto_ports_offset(ip_proto);
97 
98 	if (!data) {
99 		data = skb->data;
100 		hlen = skb_headlen(skb);
101 	}
102 
103 	if (poff >= 0) {
104 		__be32 *ports, _ports;
105 
106 		ports = __skb_header_pointer(skb, thoff + poff,
107 					     sizeof(_ports), data, hlen, &_ports);
108 		if (ports)
109 			return *ports;
110 	}
111 
112 	return 0;
113 }
114 EXPORT_SYMBOL(__skb_flow_get_ports);
115 
116 enum flow_dissect_ret {
117 	FLOW_DISSECT_RET_OUT_GOOD,
118 	FLOW_DISSECT_RET_OUT_BAD,
119 	FLOW_DISSECT_RET_OUT_PROTO_AGAIN,
120 };
121 
122 static enum flow_dissect_ret
123 __skb_flow_dissect_mpls(const struct sk_buff *skb,
124 			struct flow_dissector *flow_dissector,
125 			void *target_container, void *data, int nhoff, int hlen)
126 {
127 	struct flow_dissector_key_keyid *key_keyid;
128 	struct mpls_label *hdr, _hdr[2];
129 	u32 entry, label;
130 
131 	if (!dissector_uses_key(flow_dissector,
132 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
133 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
134 		return FLOW_DISSECT_RET_OUT_GOOD;
135 
136 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
137 				   hlen, &_hdr);
138 	if (!hdr)
139 		return FLOW_DISSECT_RET_OUT_BAD;
140 
141 	entry = ntohl(hdr[0].entry);
142 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
143 
144 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
145 		struct flow_dissector_key_mpls *key_mpls;
146 
147 		key_mpls = skb_flow_dissector_target(flow_dissector,
148 						     FLOW_DISSECTOR_KEY_MPLS,
149 						     target_container);
150 		key_mpls->mpls_label = label;
151 		key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
152 					>> MPLS_LS_TTL_SHIFT;
153 		key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
154 					>> MPLS_LS_TC_SHIFT;
155 		key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
156 					>> MPLS_LS_S_SHIFT;
157 	}
158 
159 	if (label == MPLS_LABEL_ENTROPY) {
160 		key_keyid = skb_flow_dissector_target(flow_dissector,
161 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
162 						      target_container);
163 		key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
164 	}
165 	return FLOW_DISSECT_RET_OUT_GOOD;
166 }
167 
168 static enum flow_dissect_ret
169 __skb_flow_dissect_arp(const struct sk_buff *skb,
170 		       struct flow_dissector *flow_dissector,
171 		       void *target_container, void *data, int nhoff, int hlen)
172 {
173 	struct flow_dissector_key_arp *key_arp;
174 	struct {
175 		unsigned char ar_sha[ETH_ALEN];
176 		unsigned char ar_sip[4];
177 		unsigned char ar_tha[ETH_ALEN];
178 		unsigned char ar_tip[4];
179 	} *arp_eth, _arp_eth;
180 	const struct arphdr *arp;
181 	struct arphdr _arp;
182 
183 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
184 		return FLOW_DISSECT_RET_OUT_GOOD;
185 
186 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
187 				   hlen, &_arp);
188 	if (!arp)
189 		return FLOW_DISSECT_RET_OUT_BAD;
190 
191 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
192 	    arp->ar_pro != htons(ETH_P_IP) ||
193 	    arp->ar_hln != ETH_ALEN ||
194 	    arp->ar_pln != 4 ||
195 	    (arp->ar_op != htons(ARPOP_REPLY) &&
196 	     arp->ar_op != htons(ARPOP_REQUEST)))
197 		return FLOW_DISSECT_RET_OUT_BAD;
198 
199 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
200 				       sizeof(_arp_eth), data,
201 				       hlen, &_arp_eth);
202 	if (!arp_eth)
203 		return FLOW_DISSECT_RET_OUT_BAD;
204 
205 	key_arp = skb_flow_dissector_target(flow_dissector,
206 					    FLOW_DISSECTOR_KEY_ARP,
207 					    target_container);
208 
209 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
210 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
211 
212 	/* Only store the lower byte of the opcode;
213 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
214 	 */
215 	key_arp->op = ntohs(arp->ar_op) & 0xff;
216 
217 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
218 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
219 
220 	return FLOW_DISSECT_RET_OUT_GOOD;
221 }
222 
223 static enum flow_dissect_ret
224 __skb_flow_dissect_gre(const struct sk_buff *skb,
225 		       struct flow_dissector_key_control *key_control,
226 		       struct flow_dissector *flow_dissector,
227 		       void *target_container, void *data,
228 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
229 		       unsigned int flags)
230 {
231 	struct flow_dissector_key_keyid *key_keyid;
232 	struct gre_base_hdr *hdr, _hdr;
233 	int offset = 0;
234 	u16 gre_ver;
235 
236 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
237 				   data, *p_hlen, &_hdr);
238 	if (!hdr)
239 		return FLOW_DISSECT_RET_OUT_BAD;
240 
241 	/* Only look inside GRE without routing */
242 	if (hdr->flags & GRE_ROUTING)
243 		return FLOW_DISSECT_RET_OUT_GOOD;
244 
245 	/* Only look inside GRE for version 0 and 1 */
246 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
247 	if (gre_ver > 1)
248 		return FLOW_DISSECT_RET_OUT_GOOD;
249 
250 	*p_proto = hdr->protocol;
251 	if (gre_ver) {
252 		/* Version1 must be PPTP, and check the flags */
253 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
254 			return FLOW_DISSECT_RET_OUT_GOOD;
255 	}
256 
257 	offset += sizeof(struct gre_base_hdr);
258 
259 	if (hdr->flags & GRE_CSUM)
260 		offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
261 			  sizeof(((struct gre_full_hdr *) 0)->reserved1);
262 
263 	if (hdr->flags & GRE_KEY) {
264 		const __be32 *keyid;
265 		__be32 _keyid;
266 
267 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
268 					     sizeof(_keyid),
269 					     data, *p_hlen, &_keyid);
270 		if (!keyid)
271 			return FLOW_DISSECT_RET_OUT_BAD;
272 
273 		if (dissector_uses_key(flow_dissector,
274 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
275 			key_keyid = skb_flow_dissector_target(flow_dissector,
276 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
277 							      target_container);
278 			if (gre_ver == 0)
279 				key_keyid->keyid = *keyid;
280 			else
281 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
282 		}
283 		offset += sizeof(((struct gre_full_hdr *) 0)->key);
284 	}
285 
286 	if (hdr->flags & GRE_SEQ)
287 		offset += sizeof(((struct pptp_gre_header *) 0)->seq);
288 
289 	if (gre_ver == 0) {
290 		if (*p_proto == htons(ETH_P_TEB)) {
291 			const struct ethhdr *eth;
292 			struct ethhdr _eth;
293 
294 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
295 						   sizeof(_eth),
296 						   data, *p_hlen, &_eth);
297 			if (!eth)
298 				return FLOW_DISSECT_RET_OUT_BAD;
299 			*p_proto = eth->h_proto;
300 			offset += sizeof(*eth);
301 
302 			/* Cap headers that we access via pointers at the
303 			 * end of the Ethernet header as our maximum alignment
304 			 * at that point is only 2 bytes.
305 			 */
306 			if (NET_IP_ALIGN)
307 				*p_hlen = *p_nhoff + offset;
308 		}
309 	} else { /* version 1, must be PPTP */
310 		u8 _ppp_hdr[PPP_HDRLEN];
311 		u8 *ppp_hdr;
312 
313 		if (hdr->flags & GRE_ACK)
314 			offset += sizeof(((struct pptp_gre_header *) 0)->ack);
315 
316 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
317 					       sizeof(_ppp_hdr),
318 					       data, *p_hlen, _ppp_hdr);
319 		if (!ppp_hdr)
320 			return FLOW_DISSECT_RET_OUT_BAD;
321 
322 		switch (PPP_PROTOCOL(ppp_hdr)) {
323 		case PPP_IP:
324 			*p_proto = htons(ETH_P_IP);
325 			break;
326 		case PPP_IPV6:
327 			*p_proto = htons(ETH_P_IPV6);
328 			break;
329 		default:
330 			/* Could probably catch some more like MPLS */
331 			break;
332 		}
333 
334 		offset += PPP_HDRLEN;
335 	}
336 
337 	*p_nhoff += offset;
338 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
339 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
340 		return FLOW_DISSECT_RET_OUT_GOOD;
341 
342 	return FLOW_DISSECT_RET_OUT_PROTO_AGAIN;
343 }
344 
345 /**
346  * __skb_flow_dissect - extract the flow_keys struct and return it
347  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
348  * @flow_dissector: list of keys to dissect
349  * @target_container: target structure to put dissected values into
350  * @data: raw buffer pointer to the packet, if NULL use skb->data
351  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
352  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
353  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
354  *
355  * The function will try to retrieve individual keys into target specified
356  * by flow_dissector from either the skbuff or a raw buffer specified by the
357  * rest parameters.
358  *
359  * Caller must take care of zeroing target container memory.
360  */
361 bool __skb_flow_dissect(const struct sk_buff *skb,
362 			struct flow_dissector *flow_dissector,
363 			void *target_container,
364 			void *data, __be16 proto, int nhoff, int hlen,
365 			unsigned int flags)
366 {
367 	struct flow_dissector_key_control *key_control;
368 	struct flow_dissector_key_basic *key_basic;
369 	struct flow_dissector_key_addrs *key_addrs;
370 	struct flow_dissector_key_ports *key_ports;
371 	struct flow_dissector_key_icmp *key_icmp;
372 	struct flow_dissector_key_tags *key_tags;
373 	struct flow_dissector_key_vlan *key_vlan;
374 	bool skip_vlan = false;
375 	u8 ip_proto = 0;
376 	bool ret;
377 
378 	if (!data) {
379 		data = skb->data;
380 		proto = skb_vlan_tag_present(skb) ?
381 			 skb->vlan_proto : skb->protocol;
382 		nhoff = skb_network_offset(skb);
383 		hlen = skb_headlen(skb);
384 	}
385 
386 	/* It is ensured by skb_flow_dissector_init() that control key will
387 	 * be always present.
388 	 */
389 	key_control = skb_flow_dissector_target(flow_dissector,
390 						FLOW_DISSECTOR_KEY_CONTROL,
391 						target_container);
392 
393 	/* It is ensured by skb_flow_dissector_init() that basic key will
394 	 * be always present.
395 	 */
396 	key_basic = skb_flow_dissector_target(flow_dissector,
397 					      FLOW_DISSECTOR_KEY_BASIC,
398 					      target_container);
399 
400 	if (dissector_uses_key(flow_dissector,
401 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
402 		struct ethhdr *eth = eth_hdr(skb);
403 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
404 
405 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
406 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
407 							  target_container);
408 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
409 	}
410 
411 proto_again:
412 	switch (proto) {
413 	case htons(ETH_P_IP): {
414 		const struct iphdr *iph;
415 		struct iphdr _iph;
416 ip:
417 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
418 		if (!iph || iph->ihl < 5)
419 			goto out_bad;
420 		nhoff += iph->ihl * 4;
421 
422 		ip_proto = iph->protocol;
423 
424 		if (dissector_uses_key(flow_dissector,
425 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
426 			key_addrs = skb_flow_dissector_target(flow_dissector,
427 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
428 							      target_container);
429 
430 			memcpy(&key_addrs->v4addrs, &iph->saddr,
431 			       sizeof(key_addrs->v4addrs));
432 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
433 		}
434 
435 		if (ip_is_fragment(iph)) {
436 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
437 
438 			if (iph->frag_off & htons(IP_OFFSET)) {
439 				goto out_good;
440 			} else {
441 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
442 				if (!(flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG))
443 					goto out_good;
444 			}
445 		}
446 
447 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
448 			goto out_good;
449 
450 		break;
451 	}
452 	case htons(ETH_P_IPV6): {
453 		const struct ipv6hdr *iph;
454 		struct ipv6hdr _iph;
455 
456 ipv6:
457 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
458 		if (!iph)
459 			goto out_bad;
460 
461 		ip_proto = iph->nexthdr;
462 		nhoff += sizeof(struct ipv6hdr);
463 
464 		if (dissector_uses_key(flow_dissector,
465 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
466 			key_addrs = skb_flow_dissector_target(flow_dissector,
467 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
468 							      target_container);
469 
470 			memcpy(&key_addrs->v6addrs, &iph->saddr,
471 			       sizeof(key_addrs->v6addrs));
472 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
473 		}
474 
475 		if ((dissector_uses_key(flow_dissector,
476 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
477 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
478 		    ip6_flowlabel(iph)) {
479 			__be32 flow_label = ip6_flowlabel(iph);
480 
481 			if (dissector_uses_key(flow_dissector,
482 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
483 				key_tags = skb_flow_dissector_target(flow_dissector,
484 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
485 								     target_container);
486 				key_tags->flow_label = ntohl(flow_label);
487 			}
488 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)
489 				goto out_good;
490 		}
491 
492 		if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
493 			goto out_good;
494 
495 		break;
496 	}
497 	case htons(ETH_P_8021AD):
498 	case htons(ETH_P_8021Q): {
499 		const struct vlan_hdr *vlan;
500 		struct vlan_hdr _vlan;
501 		bool vlan_tag_present = skb && skb_vlan_tag_present(skb);
502 
503 		if (vlan_tag_present)
504 			proto = skb->protocol;
505 
506 		if (!vlan_tag_present || eth_type_vlan(skb->protocol)) {
507 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
508 						    data, hlen, &_vlan);
509 			if (!vlan)
510 				goto out_bad;
511 			proto = vlan->h_vlan_encapsulated_proto;
512 			nhoff += sizeof(*vlan);
513 			if (skip_vlan)
514 				goto proto_again;
515 		}
516 
517 		skip_vlan = true;
518 		if (dissector_uses_key(flow_dissector,
519 				       FLOW_DISSECTOR_KEY_VLAN)) {
520 			key_vlan = skb_flow_dissector_target(flow_dissector,
521 							     FLOW_DISSECTOR_KEY_VLAN,
522 							     target_container);
523 
524 			if (vlan_tag_present) {
525 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
526 				key_vlan->vlan_priority =
527 					(skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
528 			} else {
529 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
530 					VLAN_VID_MASK;
531 				key_vlan->vlan_priority =
532 					(ntohs(vlan->h_vlan_TCI) &
533 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
534 			}
535 		}
536 
537 		goto proto_again;
538 	}
539 	case htons(ETH_P_PPP_SES): {
540 		struct {
541 			struct pppoe_hdr hdr;
542 			__be16 proto;
543 		} *hdr, _hdr;
544 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
545 		if (!hdr)
546 			goto out_bad;
547 		proto = hdr->proto;
548 		nhoff += PPPOE_SES_HLEN;
549 		switch (proto) {
550 		case htons(PPP_IP):
551 			goto ip;
552 		case htons(PPP_IPV6):
553 			goto ipv6;
554 		default:
555 			goto out_bad;
556 		}
557 	}
558 	case htons(ETH_P_TIPC): {
559 		struct {
560 			__be32 pre[3];
561 			__be32 srcnode;
562 		} *hdr, _hdr;
563 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
564 		if (!hdr)
565 			goto out_bad;
566 
567 		if (dissector_uses_key(flow_dissector,
568 				       FLOW_DISSECTOR_KEY_TIPC_ADDRS)) {
569 			key_addrs = skb_flow_dissector_target(flow_dissector,
570 							      FLOW_DISSECTOR_KEY_TIPC_ADDRS,
571 							      target_container);
572 			key_addrs->tipcaddrs.srcnode = hdr->srcnode;
573 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS;
574 		}
575 		goto out_good;
576 	}
577 
578 	case htons(ETH_P_MPLS_UC):
579 	case htons(ETH_P_MPLS_MC):
580 mpls:
581 		switch (__skb_flow_dissect_mpls(skb, flow_dissector,
582 						target_container, data,
583 						nhoff, hlen)) {
584 		case FLOW_DISSECT_RET_OUT_GOOD:
585 			goto out_good;
586 		case FLOW_DISSECT_RET_OUT_BAD:
587 		default:
588 			goto out_bad;
589 		}
590 	case htons(ETH_P_FCOE):
591 		if ((hlen - nhoff) < FCOE_HEADER_LEN)
592 			goto out_bad;
593 
594 		nhoff += FCOE_HEADER_LEN;
595 		goto out_good;
596 
597 	case htons(ETH_P_ARP):
598 	case htons(ETH_P_RARP):
599 		switch (__skb_flow_dissect_arp(skb, flow_dissector,
600 					       target_container, data,
601 					       nhoff, hlen)) {
602 		case FLOW_DISSECT_RET_OUT_GOOD:
603 			goto out_good;
604 		case FLOW_DISSECT_RET_OUT_BAD:
605 		default:
606 			goto out_bad;
607 		}
608 	default:
609 		goto out_bad;
610 	}
611 
612 ip_proto_again:
613 	switch (ip_proto) {
614 	case IPPROTO_GRE:
615 		switch (__skb_flow_dissect_gre(skb, key_control, flow_dissector,
616 					       target_container, data,
617 					       &proto, &nhoff, &hlen, flags)) {
618 		case FLOW_DISSECT_RET_OUT_GOOD:
619 			goto out_good;
620 		case FLOW_DISSECT_RET_OUT_BAD:
621 			goto out_bad;
622 		case FLOW_DISSECT_RET_OUT_PROTO_AGAIN:
623 			goto proto_again;
624 		}
625 	case NEXTHDR_HOP:
626 	case NEXTHDR_ROUTING:
627 	case NEXTHDR_DEST: {
628 		u8 _opthdr[2], *opthdr;
629 
630 		if (proto != htons(ETH_P_IPV6))
631 			break;
632 
633 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
634 					      data, hlen, &_opthdr);
635 		if (!opthdr)
636 			goto out_bad;
637 
638 		ip_proto = opthdr[0];
639 		nhoff += (opthdr[1] + 1) << 3;
640 
641 		goto ip_proto_again;
642 	}
643 	case NEXTHDR_FRAGMENT: {
644 		struct frag_hdr _fh, *fh;
645 
646 		if (proto != htons(ETH_P_IPV6))
647 			break;
648 
649 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
650 					  data, hlen, &_fh);
651 
652 		if (!fh)
653 			goto out_bad;
654 
655 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
656 
657 		nhoff += sizeof(_fh);
658 		ip_proto = fh->nexthdr;
659 
660 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
661 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
662 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG)
663 				goto ip_proto_again;
664 		}
665 		goto out_good;
666 	}
667 	case IPPROTO_IPIP:
668 		proto = htons(ETH_P_IP);
669 
670 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
671 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
672 			goto out_good;
673 
674 		goto ip;
675 	case IPPROTO_IPV6:
676 		proto = htons(ETH_P_IPV6);
677 
678 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
679 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
680 			goto out_good;
681 
682 		goto ipv6;
683 	case IPPROTO_MPLS:
684 		proto = htons(ETH_P_MPLS_UC);
685 		goto mpls;
686 	default:
687 		break;
688 	}
689 
690 	if (dissector_uses_key(flow_dissector,
691 			       FLOW_DISSECTOR_KEY_PORTS)) {
692 		key_ports = skb_flow_dissector_target(flow_dissector,
693 						      FLOW_DISSECTOR_KEY_PORTS,
694 						      target_container);
695 		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
696 							data, hlen);
697 	}
698 
699 	if (dissector_uses_key(flow_dissector,
700 			       FLOW_DISSECTOR_KEY_ICMP)) {
701 		key_icmp = skb_flow_dissector_target(flow_dissector,
702 						     FLOW_DISSECTOR_KEY_ICMP,
703 						     target_container);
704 		key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
705 	}
706 
707 out_good:
708 	ret = true;
709 
710 	key_control->thoff = (u16)nhoff;
711 out:
712 	key_basic->n_proto = proto;
713 	key_basic->ip_proto = ip_proto;
714 
715 	return ret;
716 
717 out_bad:
718 	ret = false;
719 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
720 	goto out;
721 }
722 EXPORT_SYMBOL(__skb_flow_dissect);
723 
724 static u32 hashrnd __read_mostly;
725 static __always_inline void __flow_hash_secret_init(void)
726 {
727 	net_get_random_once(&hashrnd, sizeof(hashrnd));
728 }
729 
730 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
731 					     u32 keyval)
732 {
733 	return jhash2(words, length, keyval);
734 }
735 
736 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
737 {
738 	const void *p = flow;
739 
740 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
741 	return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
742 }
743 
744 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
745 {
746 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
747 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
748 	BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
749 		     sizeof(*flow) - sizeof(flow->addrs));
750 
751 	switch (flow->control.addr_type) {
752 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
753 		diff -= sizeof(flow->addrs.v4addrs);
754 		break;
755 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
756 		diff -= sizeof(flow->addrs.v6addrs);
757 		break;
758 	case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
759 		diff -= sizeof(flow->addrs.tipcaddrs);
760 		break;
761 	}
762 	return (sizeof(*flow) - diff) / sizeof(u32);
763 }
764 
765 __be32 flow_get_u32_src(const struct flow_keys *flow)
766 {
767 	switch (flow->control.addr_type) {
768 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
769 		return flow->addrs.v4addrs.src;
770 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
771 		return (__force __be32)ipv6_addr_hash(
772 			&flow->addrs.v6addrs.src);
773 	case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
774 		return flow->addrs.tipcaddrs.srcnode;
775 	default:
776 		return 0;
777 	}
778 }
779 EXPORT_SYMBOL(flow_get_u32_src);
780 
781 __be32 flow_get_u32_dst(const struct flow_keys *flow)
782 {
783 	switch (flow->control.addr_type) {
784 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
785 		return flow->addrs.v4addrs.dst;
786 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
787 		return (__force __be32)ipv6_addr_hash(
788 			&flow->addrs.v6addrs.dst);
789 	default:
790 		return 0;
791 	}
792 }
793 EXPORT_SYMBOL(flow_get_u32_dst);
794 
795 static inline void __flow_hash_consistentify(struct flow_keys *keys)
796 {
797 	int addr_diff, i;
798 
799 	switch (keys->control.addr_type) {
800 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
801 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
802 			    (__force u32)keys->addrs.v4addrs.src;
803 		if ((addr_diff < 0) ||
804 		    (addr_diff == 0 &&
805 		     ((__force u16)keys->ports.dst <
806 		      (__force u16)keys->ports.src))) {
807 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
808 			swap(keys->ports.src, keys->ports.dst);
809 		}
810 		break;
811 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
812 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
813 				   &keys->addrs.v6addrs.src,
814 				   sizeof(keys->addrs.v6addrs.dst));
815 		if ((addr_diff < 0) ||
816 		    (addr_diff == 0 &&
817 		     ((__force u16)keys->ports.dst <
818 		      (__force u16)keys->ports.src))) {
819 			for (i = 0; i < 4; i++)
820 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
821 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
822 			swap(keys->ports.src, keys->ports.dst);
823 		}
824 		break;
825 	}
826 }
827 
828 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
829 {
830 	u32 hash;
831 
832 	__flow_hash_consistentify(keys);
833 
834 	hash = __flow_hash_words(flow_keys_hash_start(keys),
835 				 flow_keys_hash_length(keys), keyval);
836 	if (!hash)
837 		hash = 1;
838 
839 	return hash;
840 }
841 
842 u32 flow_hash_from_keys(struct flow_keys *keys)
843 {
844 	__flow_hash_secret_init();
845 	return __flow_hash_from_keys(keys, hashrnd);
846 }
847 EXPORT_SYMBOL(flow_hash_from_keys);
848 
849 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
850 				  struct flow_keys *keys, u32 keyval)
851 {
852 	skb_flow_dissect_flow_keys(skb, keys,
853 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
854 
855 	return __flow_hash_from_keys(keys, keyval);
856 }
857 
858 struct _flow_keys_digest_data {
859 	__be16	n_proto;
860 	u8	ip_proto;
861 	u8	padding;
862 	__be32	ports;
863 	__be32	src;
864 	__be32	dst;
865 };
866 
867 void make_flow_keys_digest(struct flow_keys_digest *digest,
868 			   const struct flow_keys *flow)
869 {
870 	struct _flow_keys_digest_data *data =
871 	    (struct _flow_keys_digest_data *)digest;
872 
873 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
874 
875 	memset(digest, 0, sizeof(*digest));
876 
877 	data->n_proto = flow->basic.n_proto;
878 	data->ip_proto = flow->basic.ip_proto;
879 	data->ports = flow->ports.ports;
880 	data->src = flow->addrs.v4addrs.src;
881 	data->dst = flow->addrs.v4addrs.dst;
882 }
883 EXPORT_SYMBOL(make_flow_keys_digest);
884 
885 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
886 
887 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
888 {
889 	struct flow_keys keys;
890 
891 	__flow_hash_secret_init();
892 
893 	memset(&keys, 0, sizeof(keys));
894 	__skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
895 			   NULL, 0, 0, 0,
896 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
897 
898 	return __flow_hash_from_keys(&keys, hashrnd);
899 }
900 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
901 
902 /**
903  * __skb_get_hash: calculate a flow hash
904  * @skb: sk_buff to calculate flow hash from
905  *
906  * This function calculates a flow hash based on src/dst addresses
907  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
908  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
909  * if hash is a canonical 4-tuple hash over transport ports.
910  */
911 void __skb_get_hash(struct sk_buff *skb)
912 {
913 	struct flow_keys keys;
914 	u32 hash;
915 
916 	__flow_hash_secret_init();
917 
918 	hash = ___skb_get_hash(skb, &keys, hashrnd);
919 
920 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
921 }
922 EXPORT_SYMBOL(__skb_get_hash);
923 
924 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
925 {
926 	struct flow_keys keys;
927 
928 	return ___skb_get_hash(skb, &keys, perturb);
929 }
930 EXPORT_SYMBOL(skb_get_hash_perturb);
931 
932 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
933 {
934 	struct flow_keys keys;
935 
936 	memset(&keys, 0, sizeof(keys));
937 
938 	memcpy(&keys.addrs.v6addrs.src, &fl6->saddr,
939 	       sizeof(keys.addrs.v6addrs.src));
940 	memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr,
941 	       sizeof(keys.addrs.v6addrs.dst));
942 	keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
943 	keys.ports.src = fl6->fl6_sport;
944 	keys.ports.dst = fl6->fl6_dport;
945 	keys.keyid.keyid = fl6->fl6_gre_key;
946 	keys.tags.flow_label = (__force u32)fl6->flowlabel;
947 	keys.basic.ip_proto = fl6->flowi6_proto;
948 
949 	__skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
950 			  flow_keys_have_l4(&keys));
951 
952 	return skb->hash;
953 }
954 EXPORT_SYMBOL(__skb_get_hash_flowi6);
955 
956 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
957 {
958 	struct flow_keys keys;
959 
960 	memset(&keys, 0, sizeof(keys));
961 
962 	keys.addrs.v4addrs.src = fl4->saddr;
963 	keys.addrs.v4addrs.dst = fl4->daddr;
964 	keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
965 	keys.ports.src = fl4->fl4_sport;
966 	keys.ports.dst = fl4->fl4_dport;
967 	keys.keyid.keyid = fl4->fl4_gre_key;
968 	keys.basic.ip_proto = fl4->flowi4_proto;
969 
970 	__skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
971 			  flow_keys_have_l4(&keys));
972 
973 	return skb->hash;
974 }
975 EXPORT_SYMBOL(__skb_get_hash_flowi4);
976 
977 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
978 		   const struct flow_keys *keys, int hlen)
979 {
980 	u32 poff = keys->control.thoff;
981 
982 	/* skip L4 headers for fragments after the first */
983 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
984 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
985 		return poff;
986 
987 	switch (keys->basic.ip_proto) {
988 	case IPPROTO_TCP: {
989 		/* access doff as u8 to avoid unaligned access */
990 		const u8 *doff;
991 		u8 _doff;
992 
993 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
994 					    data, hlen, &_doff);
995 		if (!doff)
996 			return poff;
997 
998 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
999 		break;
1000 	}
1001 	case IPPROTO_UDP:
1002 	case IPPROTO_UDPLITE:
1003 		poff += sizeof(struct udphdr);
1004 		break;
1005 	/* For the rest, we do not really care about header
1006 	 * extensions at this point for now.
1007 	 */
1008 	case IPPROTO_ICMP:
1009 		poff += sizeof(struct icmphdr);
1010 		break;
1011 	case IPPROTO_ICMPV6:
1012 		poff += sizeof(struct icmp6hdr);
1013 		break;
1014 	case IPPROTO_IGMP:
1015 		poff += sizeof(struct igmphdr);
1016 		break;
1017 	case IPPROTO_DCCP:
1018 		poff += sizeof(struct dccp_hdr);
1019 		break;
1020 	case IPPROTO_SCTP:
1021 		poff += sizeof(struct sctphdr);
1022 		break;
1023 	}
1024 
1025 	return poff;
1026 }
1027 
1028 /**
1029  * skb_get_poff - get the offset to the payload
1030  * @skb: sk_buff to get the payload offset from
1031  *
1032  * The function will get the offset to the payload as far as it could
1033  * be dissected.  The main user is currently BPF, so that we can dynamically
1034  * truncate packets without needing to push actual payload to the user
1035  * space and can analyze headers only, instead.
1036  */
1037 u32 skb_get_poff(const struct sk_buff *skb)
1038 {
1039 	struct flow_keys keys;
1040 
1041 	if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
1042 		return 0;
1043 
1044 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1045 }
1046 
1047 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1048 {
1049 	memset(keys, 0, sizeof(*keys));
1050 
1051 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1052 	    sizeof(keys->addrs.v6addrs.src));
1053 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1054 	    sizeof(keys->addrs.v6addrs.dst));
1055 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1056 	keys->ports.src = fl6->fl6_sport;
1057 	keys->ports.dst = fl6->fl6_dport;
1058 	keys->keyid.keyid = fl6->fl6_gre_key;
1059 	keys->tags.flow_label = (__force u32)fl6->flowlabel;
1060 	keys->basic.ip_proto = fl6->flowi6_proto;
1061 
1062 	return flow_hash_from_keys(keys);
1063 }
1064 EXPORT_SYMBOL(__get_hash_from_flowi6);
1065 
1066 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys)
1067 {
1068 	memset(keys, 0, sizeof(*keys));
1069 
1070 	keys->addrs.v4addrs.src = fl4->saddr;
1071 	keys->addrs.v4addrs.dst = fl4->daddr;
1072 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1073 	keys->ports.src = fl4->fl4_sport;
1074 	keys->ports.dst = fl4->fl4_dport;
1075 	keys->keyid.keyid = fl4->fl4_gre_key;
1076 	keys->basic.ip_proto = fl4->flowi4_proto;
1077 
1078 	return flow_hash_from_keys(keys);
1079 }
1080 EXPORT_SYMBOL(__get_hash_from_flowi4);
1081 
1082 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1083 	{
1084 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1085 		.offset = offsetof(struct flow_keys, control),
1086 	},
1087 	{
1088 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1089 		.offset = offsetof(struct flow_keys, basic),
1090 	},
1091 	{
1092 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1093 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1094 	},
1095 	{
1096 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1097 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1098 	},
1099 	{
1100 		.key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS,
1101 		.offset = offsetof(struct flow_keys, addrs.tipcaddrs),
1102 	},
1103 	{
1104 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1105 		.offset = offsetof(struct flow_keys, ports),
1106 	},
1107 	{
1108 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1109 		.offset = offsetof(struct flow_keys, vlan),
1110 	},
1111 	{
1112 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1113 		.offset = offsetof(struct flow_keys, tags),
1114 	},
1115 	{
1116 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1117 		.offset = offsetof(struct flow_keys, keyid),
1118 	},
1119 };
1120 
1121 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1122 	{
1123 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1124 		.offset = offsetof(struct flow_keys, control),
1125 	},
1126 	{
1127 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1128 		.offset = offsetof(struct flow_keys, basic),
1129 	},
1130 	{
1131 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1132 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1133 	},
1134 	{
1135 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1136 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1137 	},
1138 	{
1139 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1140 		.offset = offsetof(struct flow_keys, ports),
1141 	},
1142 };
1143 
1144 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
1145 	{
1146 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1147 		.offset = offsetof(struct flow_keys, control),
1148 	},
1149 	{
1150 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1151 		.offset = offsetof(struct flow_keys, basic),
1152 	},
1153 };
1154 
1155 struct flow_dissector flow_keys_dissector __read_mostly;
1156 EXPORT_SYMBOL(flow_keys_dissector);
1157 
1158 struct flow_dissector flow_keys_buf_dissector __read_mostly;
1159 
1160 static int __init init_default_flow_dissectors(void)
1161 {
1162 	skb_flow_dissector_init(&flow_keys_dissector,
1163 				flow_keys_dissector_keys,
1164 				ARRAY_SIZE(flow_keys_dissector_keys));
1165 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1166 				flow_keys_dissector_symmetric_keys,
1167 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1168 	skb_flow_dissector_init(&flow_keys_buf_dissector,
1169 				flow_keys_buf_dissector_keys,
1170 				ARRAY_SIZE(flow_keys_buf_dissector_keys));
1171 	return 0;
1172 }
1173 
1174 core_initcall(init_default_flow_dissectors);
1175