1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/kernel.h>
3 #include <linux/skbuff.h>
4 #include <linux/export.h>
5 #include <linux/ip.h>
6 #include <linux/ipv6.h>
7 #include <linux/if_vlan.h>
8 #include <net/dsa.h>
9 #include <net/dst_metadata.h>
10 #include <net/ip.h>
11 #include <net/ipv6.h>
12 #include <net/gre.h>
13 #include <net/pptp.h>
14 #include <net/tipc.h>
15 #include <linux/igmp.h>
16 #include <linux/icmp.h>
17 #include <linux/sctp.h>
18 #include <linux/dccp.h>
19 #include <linux/if_tunnel.h>
20 #include <linux/if_pppox.h>
21 #include <linux/ppp_defs.h>
22 #include <linux/stddef.h>
23 #include <linux/if_ether.h>
24 #include <linux/mpls.h>
25 #include <linux/tcp.h>
26 #include <linux/ptp_classify.h>
27 #include <net/flow_dissector.h>
28 #include <scsi/fc/fc_fcoe.h>
29 #include <uapi/linux/batadv_packet.h>
30 #include <linux/bpf.h>
31 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
32 #include <net/netfilter/nf_conntrack_core.h>
33 #include <net/netfilter/nf_conntrack_labels.h>
34 #endif
35 #include <linux/bpf-netns.h>
36 
dissector_set_key(struct flow_dissector * flow_dissector,enum flow_dissector_key_id key_id)37 static void dissector_set_key(struct flow_dissector *flow_dissector,
38 			      enum flow_dissector_key_id key_id)
39 {
40 	flow_dissector->used_keys |= (1 << key_id);
41 }
42 
skb_flow_dissector_init(struct flow_dissector * flow_dissector,const struct flow_dissector_key * key,unsigned int key_count)43 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
44 			     const struct flow_dissector_key *key,
45 			     unsigned int key_count)
46 {
47 	unsigned int i;
48 
49 	memset(flow_dissector, 0, sizeof(*flow_dissector));
50 
51 	for (i = 0; i < key_count; i++, key++) {
52 		/* User should make sure that every key target offset is within
53 		 * boundaries of unsigned short.
54 		 */
55 		BUG_ON(key->offset > USHRT_MAX);
56 		BUG_ON(dissector_uses_key(flow_dissector,
57 					  key->key_id));
58 
59 		dissector_set_key(flow_dissector, key->key_id);
60 		flow_dissector->offset[key->key_id] = key->offset;
61 	}
62 
63 	/* Ensure that the dissector always includes control and basic key.
64 	 * That way we are able to avoid handling lack of these in fast path.
65 	 */
66 	BUG_ON(!dissector_uses_key(flow_dissector,
67 				   FLOW_DISSECTOR_KEY_CONTROL));
68 	BUG_ON(!dissector_uses_key(flow_dissector,
69 				   FLOW_DISSECTOR_KEY_BASIC));
70 }
71 EXPORT_SYMBOL(skb_flow_dissector_init);
72 
73 #ifdef CONFIG_BPF_SYSCALL
flow_dissector_bpf_prog_attach_check(struct net * net,struct bpf_prog * prog)74 int flow_dissector_bpf_prog_attach_check(struct net *net,
75 					 struct bpf_prog *prog)
76 {
77 	enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
78 
79 	if (net == &init_net) {
80 		/* BPF flow dissector in the root namespace overrides
81 		 * any per-net-namespace one. When attaching to root,
82 		 * make sure we don't have any BPF program attached
83 		 * to the non-root namespaces.
84 		 */
85 		struct net *ns;
86 
87 		for_each_net(ns) {
88 			if (ns == &init_net)
89 				continue;
90 			if (rcu_access_pointer(ns->bpf.run_array[type]))
91 				return -EEXIST;
92 		}
93 	} else {
94 		/* Make sure root flow dissector is not attached
95 		 * when attaching to the non-root namespace.
96 		 */
97 		if (rcu_access_pointer(init_net.bpf.run_array[type]))
98 			return -EEXIST;
99 	}
100 
101 	return 0;
102 }
103 #endif /* CONFIG_BPF_SYSCALL */
104 
105 /**
106  * __skb_flow_get_ports - extract the upper layer ports and return them
107  * @skb: sk_buff to extract the ports from
108  * @thoff: transport header offset
109  * @ip_proto: protocol for which to get port offset
110  * @data: raw buffer pointer to the packet, if NULL use skb->data
111  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
112  *
113  * The function will try to retrieve the ports at offset thoff + poff where poff
114  * is the protocol port offset returned from proto_ports_offset
115  */
__skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto,const void * data,int hlen)116 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
117 			    const void *data, int hlen)
118 {
119 	int poff = proto_ports_offset(ip_proto);
120 
121 	if (!data) {
122 		data = skb->data;
123 		hlen = skb_headlen(skb);
124 	}
125 
126 	if (poff >= 0) {
127 		__be32 *ports, _ports;
128 
129 		ports = __skb_header_pointer(skb, thoff + poff,
130 					     sizeof(_ports), data, hlen, &_ports);
131 		if (ports)
132 			return *ports;
133 	}
134 
135 	return 0;
136 }
137 EXPORT_SYMBOL(__skb_flow_get_ports);
138 
icmp_has_id(u8 type)139 static bool icmp_has_id(u8 type)
140 {
141 	switch (type) {
142 	case ICMP_ECHO:
143 	case ICMP_ECHOREPLY:
144 	case ICMP_TIMESTAMP:
145 	case ICMP_TIMESTAMPREPLY:
146 	case ICMPV6_ECHO_REQUEST:
147 	case ICMPV6_ECHO_REPLY:
148 		return true;
149 	}
150 
151 	return false;
152 }
153 
154 /**
155  * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
156  * @skb: sk_buff to extract from
157  * @key_icmp: struct flow_dissector_key_icmp to fill
158  * @data: raw buffer pointer to the packet
159  * @thoff: offset to extract at
160  * @hlen: packet header length
161  */
skb_flow_get_icmp_tci(const struct sk_buff * skb,struct flow_dissector_key_icmp * key_icmp,const void * data,int thoff,int hlen)162 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
163 			   struct flow_dissector_key_icmp *key_icmp,
164 			   const void *data, int thoff, int hlen)
165 {
166 	struct icmphdr *ih, _ih;
167 
168 	ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
169 	if (!ih)
170 		return;
171 
172 	key_icmp->type = ih->type;
173 	key_icmp->code = ih->code;
174 
175 	/* As we use 0 to signal that the Id field is not present,
176 	 * avoid confusion with packets without such field
177 	 */
178 	if (icmp_has_id(ih->type))
179 		key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
180 	else
181 		key_icmp->id = 0;
182 }
183 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
184 
185 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
186  * using skb_flow_get_icmp_tci().
187  */
__skb_flow_dissect_icmp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int thoff,int hlen)188 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
189 				    struct flow_dissector *flow_dissector,
190 				    void *target_container, const void *data,
191 				    int thoff, int hlen)
192 {
193 	struct flow_dissector_key_icmp *key_icmp;
194 
195 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
196 		return;
197 
198 	key_icmp = skb_flow_dissector_target(flow_dissector,
199 					     FLOW_DISSECTOR_KEY_ICMP,
200 					     target_container);
201 
202 	skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
203 }
204 
skb_flow_dissect_meta(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)205 void skb_flow_dissect_meta(const struct sk_buff *skb,
206 			   struct flow_dissector *flow_dissector,
207 			   void *target_container)
208 {
209 	struct flow_dissector_key_meta *meta;
210 
211 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
212 		return;
213 
214 	meta = skb_flow_dissector_target(flow_dissector,
215 					 FLOW_DISSECTOR_KEY_META,
216 					 target_container);
217 	meta->ingress_ifindex = skb->skb_iif;
218 }
219 EXPORT_SYMBOL(skb_flow_dissect_meta);
220 
221 static void
skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,struct flow_dissector * flow_dissector,void * target_container)222 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
223 				   struct flow_dissector *flow_dissector,
224 				   void *target_container)
225 {
226 	struct flow_dissector_key_control *ctrl;
227 
228 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
229 		return;
230 
231 	ctrl = skb_flow_dissector_target(flow_dissector,
232 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
233 					 target_container);
234 	ctrl->addr_type = type;
235 }
236 
237 void
skb_flow_dissect_ct(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,u16 * ctinfo_map,size_t mapsize,bool post_ct,u16 zone)238 skb_flow_dissect_ct(const struct sk_buff *skb,
239 		    struct flow_dissector *flow_dissector,
240 		    void *target_container, u16 *ctinfo_map,
241 		    size_t mapsize, bool post_ct, u16 zone)
242 {
243 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
244 	struct flow_dissector_key_ct *key;
245 	enum ip_conntrack_info ctinfo;
246 	struct nf_conn_labels *cl;
247 	struct nf_conn *ct;
248 
249 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
250 		return;
251 
252 	ct = nf_ct_get(skb, &ctinfo);
253 	if (!ct && !post_ct)
254 		return;
255 
256 	key = skb_flow_dissector_target(flow_dissector,
257 					FLOW_DISSECTOR_KEY_CT,
258 					target_container);
259 
260 	if (!ct) {
261 		key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
262 				TCA_FLOWER_KEY_CT_FLAGS_INVALID;
263 		key->ct_zone = zone;
264 		return;
265 	}
266 
267 	if (ctinfo < mapsize)
268 		key->ct_state = ctinfo_map[ctinfo];
269 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
270 	key->ct_zone = ct->zone.id;
271 #endif
272 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
273 	key->ct_mark = ct->mark;
274 #endif
275 
276 	cl = nf_ct_labels_find(ct);
277 	if (cl)
278 		memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
279 #endif /* CONFIG_NF_CONNTRACK */
280 }
281 EXPORT_SYMBOL(skb_flow_dissect_ct);
282 
283 void
skb_flow_dissect_tunnel_info(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)284 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
285 			     struct flow_dissector *flow_dissector,
286 			     void *target_container)
287 {
288 	struct ip_tunnel_info *info;
289 	struct ip_tunnel_key *key;
290 
291 	/* A quick check to see if there might be something to do. */
292 	if (!dissector_uses_key(flow_dissector,
293 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
294 	    !dissector_uses_key(flow_dissector,
295 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
296 	    !dissector_uses_key(flow_dissector,
297 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
298 	    !dissector_uses_key(flow_dissector,
299 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
300 	    !dissector_uses_key(flow_dissector,
301 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
302 	    !dissector_uses_key(flow_dissector,
303 				FLOW_DISSECTOR_KEY_ENC_IP) &&
304 	    !dissector_uses_key(flow_dissector,
305 				FLOW_DISSECTOR_KEY_ENC_OPTS))
306 		return;
307 
308 	info = skb_tunnel_info(skb);
309 	if (!info)
310 		return;
311 
312 	key = &info->key;
313 
314 	switch (ip_tunnel_info_af(info)) {
315 	case AF_INET:
316 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
317 						   flow_dissector,
318 						   target_container);
319 		if (dissector_uses_key(flow_dissector,
320 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
321 			struct flow_dissector_key_ipv4_addrs *ipv4;
322 
323 			ipv4 = skb_flow_dissector_target(flow_dissector,
324 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
325 							 target_container);
326 			ipv4->src = key->u.ipv4.src;
327 			ipv4->dst = key->u.ipv4.dst;
328 		}
329 		break;
330 	case AF_INET6:
331 		skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
332 						   flow_dissector,
333 						   target_container);
334 		if (dissector_uses_key(flow_dissector,
335 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
336 			struct flow_dissector_key_ipv6_addrs *ipv6;
337 
338 			ipv6 = skb_flow_dissector_target(flow_dissector,
339 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
340 							 target_container);
341 			ipv6->src = key->u.ipv6.src;
342 			ipv6->dst = key->u.ipv6.dst;
343 		}
344 		break;
345 	}
346 
347 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
348 		struct flow_dissector_key_keyid *keyid;
349 
350 		keyid = skb_flow_dissector_target(flow_dissector,
351 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
352 						  target_container);
353 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
354 	}
355 
356 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
357 		struct flow_dissector_key_ports *tp;
358 
359 		tp = skb_flow_dissector_target(flow_dissector,
360 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
361 					       target_container);
362 		tp->src = key->tp_src;
363 		tp->dst = key->tp_dst;
364 	}
365 
366 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
367 		struct flow_dissector_key_ip *ip;
368 
369 		ip = skb_flow_dissector_target(flow_dissector,
370 					       FLOW_DISSECTOR_KEY_ENC_IP,
371 					       target_container);
372 		ip->tos = key->tos;
373 		ip->ttl = key->ttl;
374 	}
375 
376 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
377 		struct flow_dissector_key_enc_opts *enc_opt;
378 
379 		enc_opt = skb_flow_dissector_target(flow_dissector,
380 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
381 						    target_container);
382 
383 		if (info->options_len) {
384 			enc_opt->len = info->options_len;
385 			ip_tunnel_info_opts_get(enc_opt->data, info);
386 			enc_opt->dst_opt_type = info->key.tun_flags &
387 						TUNNEL_OPTIONS_PRESENT;
388 		}
389 	}
390 }
391 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
392 
skb_flow_dissect_hash(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)393 void skb_flow_dissect_hash(const struct sk_buff *skb,
394 			   struct flow_dissector *flow_dissector,
395 			   void *target_container)
396 {
397 	struct flow_dissector_key_hash *key;
398 
399 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
400 		return;
401 
402 	key = skb_flow_dissector_target(flow_dissector,
403 					FLOW_DISSECTOR_KEY_HASH,
404 					target_container);
405 
406 	key->hash = skb_get_hash_raw(skb);
407 }
408 EXPORT_SYMBOL(skb_flow_dissect_hash);
409 
410 static enum flow_dissect_ret
__skb_flow_dissect_mpls(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen,int lse_index,bool * entropy_label)411 __skb_flow_dissect_mpls(const struct sk_buff *skb,
412 			struct flow_dissector *flow_dissector,
413 			void *target_container, const void *data, int nhoff,
414 			int hlen, int lse_index, bool *entropy_label)
415 {
416 	struct mpls_label *hdr, _hdr;
417 	u32 entry, label, bos;
418 
419 	if (!dissector_uses_key(flow_dissector,
420 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
421 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
422 		return FLOW_DISSECT_RET_OUT_GOOD;
423 
424 	if (lse_index >= FLOW_DIS_MPLS_MAX)
425 		return FLOW_DISSECT_RET_OUT_GOOD;
426 
427 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
428 				   hlen, &_hdr);
429 	if (!hdr)
430 		return FLOW_DISSECT_RET_OUT_BAD;
431 
432 	entry = ntohl(hdr->entry);
433 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
434 	bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
435 
436 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
437 		struct flow_dissector_key_mpls *key_mpls;
438 		struct flow_dissector_mpls_lse *lse;
439 
440 		key_mpls = skb_flow_dissector_target(flow_dissector,
441 						     FLOW_DISSECTOR_KEY_MPLS,
442 						     target_container);
443 		lse = &key_mpls->ls[lse_index];
444 
445 		lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
446 		lse->mpls_bos = bos;
447 		lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
448 		lse->mpls_label = label;
449 		dissector_set_mpls_lse(key_mpls, lse_index);
450 	}
451 
452 	if (*entropy_label &&
453 	    dissector_uses_key(flow_dissector,
454 			       FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
455 		struct flow_dissector_key_keyid *key_keyid;
456 
457 		key_keyid = skb_flow_dissector_target(flow_dissector,
458 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
459 						      target_container);
460 		key_keyid->keyid = cpu_to_be32(label);
461 	}
462 
463 	*entropy_label = label == MPLS_LABEL_ENTROPY;
464 
465 	return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
466 }
467 
468 static enum flow_dissect_ret
__skb_flow_dissect_arp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)469 __skb_flow_dissect_arp(const struct sk_buff *skb,
470 		       struct flow_dissector *flow_dissector,
471 		       void *target_container, const void *data,
472 		       int nhoff, int hlen)
473 {
474 	struct flow_dissector_key_arp *key_arp;
475 	struct {
476 		unsigned char ar_sha[ETH_ALEN];
477 		unsigned char ar_sip[4];
478 		unsigned char ar_tha[ETH_ALEN];
479 		unsigned char ar_tip[4];
480 	} *arp_eth, _arp_eth;
481 	const struct arphdr *arp;
482 	struct arphdr _arp;
483 
484 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
485 		return FLOW_DISSECT_RET_OUT_GOOD;
486 
487 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
488 				   hlen, &_arp);
489 	if (!arp)
490 		return FLOW_DISSECT_RET_OUT_BAD;
491 
492 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
493 	    arp->ar_pro != htons(ETH_P_IP) ||
494 	    arp->ar_hln != ETH_ALEN ||
495 	    arp->ar_pln != 4 ||
496 	    (arp->ar_op != htons(ARPOP_REPLY) &&
497 	     arp->ar_op != htons(ARPOP_REQUEST)))
498 		return FLOW_DISSECT_RET_OUT_BAD;
499 
500 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
501 				       sizeof(_arp_eth), data,
502 				       hlen, &_arp_eth);
503 	if (!arp_eth)
504 		return FLOW_DISSECT_RET_OUT_BAD;
505 
506 	key_arp = skb_flow_dissector_target(flow_dissector,
507 					    FLOW_DISSECTOR_KEY_ARP,
508 					    target_container);
509 
510 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
511 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
512 
513 	/* Only store the lower byte of the opcode;
514 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
515 	 */
516 	key_arp->op = ntohs(arp->ar_op) & 0xff;
517 
518 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
519 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
520 
521 	return FLOW_DISSECT_RET_OUT_GOOD;
522 }
523 
524 static enum flow_dissect_ret
__skb_flow_dissect_gre(const struct sk_buff * skb,struct flow_dissector_key_control * key_control,struct flow_dissector * flow_dissector,void * target_container,const void * data,__be16 * p_proto,int * p_nhoff,int * p_hlen,unsigned int flags)525 __skb_flow_dissect_gre(const struct sk_buff *skb,
526 		       struct flow_dissector_key_control *key_control,
527 		       struct flow_dissector *flow_dissector,
528 		       void *target_container, const void *data,
529 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
530 		       unsigned int flags)
531 {
532 	struct flow_dissector_key_keyid *key_keyid;
533 	struct gre_base_hdr *hdr, _hdr;
534 	int offset = 0;
535 	u16 gre_ver;
536 
537 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
538 				   data, *p_hlen, &_hdr);
539 	if (!hdr)
540 		return FLOW_DISSECT_RET_OUT_BAD;
541 
542 	/* Only look inside GRE without routing */
543 	if (hdr->flags & GRE_ROUTING)
544 		return FLOW_DISSECT_RET_OUT_GOOD;
545 
546 	/* Only look inside GRE for version 0 and 1 */
547 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
548 	if (gre_ver > 1)
549 		return FLOW_DISSECT_RET_OUT_GOOD;
550 
551 	*p_proto = hdr->protocol;
552 	if (gre_ver) {
553 		/* Version1 must be PPTP, and check the flags */
554 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
555 			return FLOW_DISSECT_RET_OUT_GOOD;
556 	}
557 
558 	offset += sizeof(struct gre_base_hdr);
559 
560 	if (hdr->flags & GRE_CSUM)
561 		offset += sizeof_field(struct gre_full_hdr, csum) +
562 			  sizeof_field(struct gre_full_hdr, reserved1);
563 
564 	if (hdr->flags & GRE_KEY) {
565 		const __be32 *keyid;
566 		__be32 _keyid;
567 
568 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
569 					     sizeof(_keyid),
570 					     data, *p_hlen, &_keyid);
571 		if (!keyid)
572 			return FLOW_DISSECT_RET_OUT_BAD;
573 
574 		if (dissector_uses_key(flow_dissector,
575 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
576 			key_keyid = skb_flow_dissector_target(flow_dissector,
577 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
578 							      target_container);
579 			if (gre_ver == 0)
580 				key_keyid->keyid = *keyid;
581 			else
582 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
583 		}
584 		offset += sizeof_field(struct gre_full_hdr, key);
585 	}
586 
587 	if (hdr->flags & GRE_SEQ)
588 		offset += sizeof_field(struct pptp_gre_header, seq);
589 
590 	if (gre_ver == 0) {
591 		if (*p_proto == htons(ETH_P_TEB)) {
592 			const struct ethhdr *eth;
593 			struct ethhdr _eth;
594 
595 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
596 						   sizeof(_eth),
597 						   data, *p_hlen, &_eth);
598 			if (!eth)
599 				return FLOW_DISSECT_RET_OUT_BAD;
600 			*p_proto = eth->h_proto;
601 			offset += sizeof(*eth);
602 
603 			/* Cap headers that we access via pointers at the
604 			 * end of the Ethernet header as our maximum alignment
605 			 * at that point is only 2 bytes.
606 			 */
607 			if (NET_IP_ALIGN)
608 				*p_hlen = *p_nhoff + offset;
609 		}
610 	} else { /* version 1, must be PPTP */
611 		u8 _ppp_hdr[PPP_HDRLEN];
612 		u8 *ppp_hdr;
613 
614 		if (hdr->flags & GRE_ACK)
615 			offset += sizeof_field(struct pptp_gre_header, ack);
616 
617 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
618 					       sizeof(_ppp_hdr),
619 					       data, *p_hlen, _ppp_hdr);
620 		if (!ppp_hdr)
621 			return FLOW_DISSECT_RET_OUT_BAD;
622 
623 		switch (PPP_PROTOCOL(ppp_hdr)) {
624 		case PPP_IP:
625 			*p_proto = htons(ETH_P_IP);
626 			break;
627 		case PPP_IPV6:
628 			*p_proto = htons(ETH_P_IPV6);
629 			break;
630 		default:
631 			/* Could probably catch some more like MPLS */
632 			break;
633 		}
634 
635 		offset += PPP_HDRLEN;
636 	}
637 
638 	*p_nhoff += offset;
639 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
640 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
641 		return FLOW_DISSECT_RET_OUT_GOOD;
642 
643 	return FLOW_DISSECT_RET_PROTO_AGAIN;
644 }
645 
646 /**
647  * __skb_flow_dissect_batadv() - dissect batman-adv header
648  * @skb: sk_buff to with the batman-adv header
649  * @key_control: flow dissectors control key
650  * @data: raw buffer pointer to the packet, if NULL use skb->data
651  * @p_proto: pointer used to update the protocol to process next
652  * @p_nhoff: pointer used to update inner network header offset
653  * @hlen: packet header length
654  * @flags: any combination of FLOW_DISSECTOR_F_*
655  *
656  * ETH_P_BATMAN packets are tried to be dissected. Only
657  * &struct batadv_unicast packets are actually processed because they contain an
658  * inner ethernet header and are usually followed by actual network header. This
659  * allows the flow dissector to continue processing the packet.
660  *
661  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
662  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
663  *  otherwise FLOW_DISSECT_RET_OUT_BAD
664  */
665 static enum flow_dissect_ret
__skb_flow_dissect_batadv(const struct sk_buff * skb,struct flow_dissector_key_control * key_control,const void * data,__be16 * p_proto,int * p_nhoff,int hlen,unsigned int flags)666 __skb_flow_dissect_batadv(const struct sk_buff *skb,
667 			  struct flow_dissector_key_control *key_control,
668 			  const void *data, __be16 *p_proto, int *p_nhoff,
669 			  int hlen, unsigned int flags)
670 {
671 	struct {
672 		struct batadv_unicast_packet batadv_unicast;
673 		struct ethhdr eth;
674 	} *hdr, _hdr;
675 
676 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
677 				   &_hdr);
678 	if (!hdr)
679 		return FLOW_DISSECT_RET_OUT_BAD;
680 
681 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
682 		return FLOW_DISSECT_RET_OUT_BAD;
683 
684 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
685 		return FLOW_DISSECT_RET_OUT_BAD;
686 
687 	*p_proto = hdr->eth.h_proto;
688 	*p_nhoff += sizeof(*hdr);
689 
690 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
691 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
692 		return FLOW_DISSECT_RET_OUT_GOOD;
693 
694 	return FLOW_DISSECT_RET_PROTO_AGAIN;
695 }
696 
697 static void
__skb_flow_dissect_tcp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int thoff,int hlen)698 __skb_flow_dissect_tcp(const struct sk_buff *skb,
699 		       struct flow_dissector *flow_dissector,
700 		       void *target_container, const void *data,
701 		       int thoff, int hlen)
702 {
703 	struct flow_dissector_key_tcp *key_tcp;
704 	struct tcphdr *th, _th;
705 
706 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
707 		return;
708 
709 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
710 	if (!th)
711 		return;
712 
713 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
714 		return;
715 
716 	key_tcp = skb_flow_dissector_target(flow_dissector,
717 					    FLOW_DISSECTOR_KEY_TCP,
718 					    target_container);
719 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
720 }
721 
722 static void
__skb_flow_dissect_ports(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,u8 ip_proto,int hlen)723 __skb_flow_dissect_ports(const struct sk_buff *skb,
724 			 struct flow_dissector *flow_dissector,
725 			 void *target_container, const void *data,
726 			 int nhoff, u8 ip_proto, int hlen)
727 {
728 	enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
729 	struct flow_dissector_key_ports *key_ports;
730 
731 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
732 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
733 	else if (dissector_uses_key(flow_dissector,
734 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
735 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
736 
737 	if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
738 		return;
739 
740 	key_ports = skb_flow_dissector_target(flow_dissector,
741 					      dissector_ports,
742 					      target_container);
743 	key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
744 						data, hlen);
745 }
746 
747 static void
__skb_flow_dissect_ipv4(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,const struct iphdr * iph)748 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
749 			struct flow_dissector *flow_dissector,
750 			void *target_container, const void *data,
751 			const struct iphdr *iph)
752 {
753 	struct flow_dissector_key_ip *key_ip;
754 
755 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
756 		return;
757 
758 	key_ip = skb_flow_dissector_target(flow_dissector,
759 					   FLOW_DISSECTOR_KEY_IP,
760 					   target_container);
761 	key_ip->tos = iph->tos;
762 	key_ip->ttl = iph->ttl;
763 }
764 
765 static void
__skb_flow_dissect_ipv6(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,const struct ipv6hdr * iph)766 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
767 			struct flow_dissector *flow_dissector,
768 			void *target_container, const void *data,
769 			const struct ipv6hdr *iph)
770 {
771 	struct flow_dissector_key_ip *key_ip;
772 
773 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
774 		return;
775 
776 	key_ip = skb_flow_dissector_target(flow_dissector,
777 					   FLOW_DISSECTOR_KEY_IP,
778 					   target_container);
779 	key_ip->tos = ipv6_get_dsfield(iph);
780 	key_ip->ttl = iph->hop_limit;
781 }
782 
783 /* Maximum number of protocol headers that can be parsed in
784  * __skb_flow_dissect
785  */
786 #define MAX_FLOW_DISSECT_HDRS	15
787 
skb_flow_dissect_allowed(int * num_hdrs)788 static bool skb_flow_dissect_allowed(int *num_hdrs)
789 {
790 	++*num_hdrs;
791 
792 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
793 }
794 
__skb_flow_bpf_to_target(const struct bpf_flow_keys * flow_keys,struct flow_dissector * flow_dissector,void * target_container)795 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
796 				     struct flow_dissector *flow_dissector,
797 				     void *target_container)
798 {
799 	struct flow_dissector_key_ports *key_ports = NULL;
800 	struct flow_dissector_key_control *key_control;
801 	struct flow_dissector_key_basic *key_basic;
802 	struct flow_dissector_key_addrs *key_addrs;
803 	struct flow_dissector_key_tags *key_tags;
804 
805 	key_control = skb_flow_dissector_target(flow_dissector,
806 						FLOW_DISSECTOR_KEY_CONTROL,
807 						target_container);
808 	key_control->thoff = flow_keys->thoff;
809 	if (flow_keys->is_frag)
810 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
811 	if (flow_keys->is_first_frag)
812 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
813 	if (flow_keys->is_encap)
814 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
815 
816 	key_basic = skb_flow_dissector_target(flow_dissector,
817 					      FLOW_DISSECTOR_KEY_BASIC,
818 					      target_container);
819 	key_basic->n_proto = flow_keys->n_proto;
820 	key_basic->ip_proto = flow_keys->ip_proto;
821 
822 	if (flow_keys->addr_proto == ETH_P_IP &&
823 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
824 		key_addrs = skb_flow_dissector_target(flow_dissector,
825 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
826 						      target_container);
827 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
828 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
829 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
830 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
831 		   dissector_uses_key(flow_dissector,
832 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
833 		key_addrs = skb_flow_dissector_target(flow_dissector,
834 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
835 						      target_container);
836 		memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
837 		       sizeof(key_addrs->v6addrs.src));
838 		memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
839 		       sizeof(key_addrs->v6addrs.dst));
840 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
841 	}
842 
843 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
844 		key_ports = skb_flow_dissector_target(flow_dissector,
845 						      FLOW_DISSECTOR_KEY_PORTS,
846 						      target_container);
847 	else if (dissector_uses_key(flow_dissector,
848 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
849 		key_ports = skb_flow_dissector_target(flow_dissector,
850 						      FLOW_DISSECTOR_KEY_PORTS_RANGE,
851 						      target_container);
852 
853 	if (key_ports) {
854 		key_ports->src = flow_keys->sport;
855 		key_ports->dst = flow_keys->dport;
856 	}
857 
858 	if (dissector_uses_key(flow_dissector,
859 			       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
860 		key_tags = skb_flow_dissector_target(flow_dissector,
861 						     FLOW_DISSECTOR_KEY_FLOW_LABEL,
862 						     target_container);
863 		key_tags->flow_label = ntohl(flow_keys->flow_label);
864 	}
865 }
866 
bpf_flow_dissect(struct bpf_prog * prog,struct bpf_flow_dissector * ctx,__be16 proto,int nhoff,int hlen,unsigned int flags)867 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
868 		      __be16 proto, int nhoff, int hlen, unsigned int flags)
869 {
870 	struct bpf_flow_keys *flow_keys = ctx->flow_keys;
871 	u32 result;
872 
873 	/* Pass parameters to the BPF program */
874 	memset(flow_keys, 0, sizeof(*flow_keys));
875 	flow_keys->n_proto = proto;
876 	flow_keys->nhoff = nhoff;
877 	flow_keys->thoff = flow_keys->nhoff;
878 
879 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
880 		     (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
881 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
882 		     (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
883 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
884 		     (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
885 	flow_keys->flags = flags;
886 
887 	result = bpf_prog_run_pin_on_cpu(prog, ctx);
888 
889 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
890 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
891 				   flow_keys->nhoff, hlen);
892 
893 	return result == BPF_OK;
894 }
895 
896 /**
897  * __skb_flow_dissect - extract the flow_keys struct and return it
898  * @net: associated network namespace, derived from @skb if NULL
899  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
900  * @flow_dissector: list of keys to dissect
901  * @target_container: target structure to put dissected values into
902  * @data: raw buffer pointer to the packet, if NULL use skb->data
903  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
904  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
905  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
906  * @flags: flags that control the dissection process, e.g.
907  *         FLOW_DISSECTOR_F_STOP_AT_ENCAP.
908  *
909  * The function will try to retrieve individual keys into target specified
910  * by flow_dissector from either the skbuff or a raw buffer specified by the
911  * rest parameters.
912  *
913  * Caller must take care of zeroing target container memory.
914  */
__skb_flow_dissect(const struct net * net,const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,__be16 proto,int nhoff,int hlen,unsigned int flags)915 bool __skb_flow_dissect(const struct net *net,
916 			const struct sk_buff *skb,
917 			struct flow_dissector *flow_dissector,
918 			void *target_container, const void *data,
919 			__be16 proto, int nhoff, int hlen, unsigned int flags)
920 {
921 	struct flow_dissector_key_control *key_control;
922 	struct flow_dissector_key_basic *key_basic;
923 	struct flow_dissector_key_addrs *key_addrs;
924 	struct flow_dissector_key_tags *key_tags;
925 	struct flow_dissector_key_vlan *key_vlan;
926 	enum flow_dissect_ret fdret;
927 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
928 	bool mpls_el = false;
929 	int mpls_lse = 0;
930 	int num_hdrs = 0;
931 	u8 ip_proto = 0;
932 	bool ret;
933 
934 	if (!data) {
935 		data = skb->data;
936 		proto = skb_vlan_tag_present(skb) ?
937 			 skb->vlan_proto : skb->protocol;
938 		nhoff = skb_network_offset(skb);
939 		hlen = skb_headlen(skb);
940 #if IS_ENABLED(CONFIG_NET_DSA)
941 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
942 			     proto == htons(ETH_P_XDSA))) {
943 			const struct dsa_device_ops *ops;
944 			int offset = 0;
945 
946 			ops = skb->dev->dsa_ptr->tag_ops;
947 			/* Only DSA header taggers break flow dissection */
948 			if (ops->needed_headroom) {
949 				if (ops->flow_dissect)
950 					ops->flow_dissect(skb, &proto, &offset);
951 				else
952 					dsa_tag_generic_flow_dissect(skb,
953 								     &proto,
954 								     &offset);
955 				hlen -= offset;
956 				nhoff += offset;
957 			}
958 		}
959 #endif
960 	}
961 
962 	/* It is ensured by skb_flow_dissector_init() that control key will
963 	 * be always present.
964 	 */
965 	key_control = skb_flow_dissector_target(flow_dissector,
966 						FLOW_DISSECTOR_KEY_CONTROL,
967 						target_container);
968 
969 	/* It is ensured by skb_flow_dissector_init() that basic key will
970 	 * be always present.
971 	 */
972 	key_basic = skb_flow_dissector_target(flow_dissector,
973 					      FLOW_DISSECTOR_KEY_BASIC,
974 					      target_container);
975 
976 	if (skb) {
977 		if (!net) {
978 			if (skb->dev)
979 				net = dev_net(skb->dev);
980 			else if (skb->sk)
981 				net = sock_net(skb->sk);
982 		}
983 	}
984 
985 	WARN_ON_ONCE(!net);
986 	if (net) {
987 		enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
988 		struct bpf_prog_array *run_array;
989 
990 		rcu_read_lock();
991 		run_array = rcu_dereference(init_net.bpf.run_array[type]);
992 		if (!run_array)
993 			run_array = rcu_dereference(net->bpf.run_array[type]);
994 
995 		if (run_array) {
996 			struct bpf_flow_keys flow_keys;
997 			struct bpf_flow_dissector ctx = {
998 				.flow_keys = &flow_keys,
999 				.data = data,
1000 				.data_end = data + hlen,
1001 			};
1002 			__be16 n_proto = proto;
1003 			struct bpf_prog *prog;
1004 
1005 			if (skb) {
1006 				ctx.skb = skb;
1007 				/* we can't use 'proto' in the skb case
1008 				 * because it might be set to skb->vlan_proto
1009 				 * which has been pulled from the data
1010 				 */
1011 				n_proto = skb->protocol;
1012 			}
1013 
1014 			prog = READ_ONCE(run_array->items[0].prog);
1015 			ret = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1016 					       hlen, flags);
1017 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1018 						 target_container);
1019 			rcu_read_unlock();
1020 			return ret;
1021 		}
1022 		rcu_read_unlock();
1023 	}
1024 
1025 	if (dissector_uses_key(flow_dissector,
1026 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1027 		struct ethhdr *eth = eth_hdr(skb);
1028 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
1029 
1030 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1031 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
1032 							  target_container);
1033 		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
1034 	}
1035 
1036 proto_again:
1037 	fdret = FLOW_DISSECT_RET_CONTINUE;
1038 
1039 	switch (proto) {
1040 	case htons(ETH_P_IP): {
1041 		const struct iphdr *iph;
1042 		struct iphdr _iph;
1043 
1044 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1045 		if (!iph || iph->ihl < 5) {
1046 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1047 			break;
1048 		}
1049 
1050 		nhoff += iph->ihl * 4;
1051 
1052 		ip_proto = iph->protocol;
1053 
1054 		if (dissector_uses_key(flow_dissector,
1055 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1056 			key_addrs = skb_flow_dissector_target(flow_dissector,
1057 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1058 							      target_container);
1059 
1060 			memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1061 			       sizeof(key_addrs->v4addrs.src));
1062 			memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1063 			       sizeof(key_addrs->v4addrs.dst));
1064 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1065 		}
1066 
1067 		__skb_flow_dissect_ipv4(skb, flow_dissector,
1068 					target_container, data, iph);
1069 
1070 		if (ip_is_fragment(iph)) {
1071 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1072 
1073 			if (iph->frag_off & htons(IP_OFFSET)) {
1074 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1075 				break;
1076 			} else {
1077 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
1078 				if (!(flags &
1079 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1080 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
1081 					break;
1082 				}
1083 			}
1084 		}
1085 
1086 		break;
1087 	}
1088 	case htons(ETH_P_IPV6): {
1089 		const struct ipv6hdr *iph;
1090 		struct ipv6hdr _iph;
1091 
1092 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1093 		if (!iph) {
1094 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1095 			break;
1096 		}
1097 
1098 		ip_proto = iph->nexthdr;
1099 		nhoff += sizeof(struct ipv6hdr);
1100 
1101 		if (dissector_uses_key(flow_dissector,
1102 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1103 			key_addrs = skb_flow_dissector_target(flow_dissector,
1104 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1105 							      target_container);
1106 
1107 			memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1108 			       sizeof(key_addrs->v6addrs.src));
1109 			memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1110 			       sizeof(key_addrs->v6addrs.dst));
1111 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1112 		}
1113 
1114 		if ((dissector_uses_key(flow_dissector,
1115 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1116 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1117 		    ip6_flowlabel(iph)) {
1118 			__be32 flow_label = ip6_flowlabel(iph);
1119 
1120 			if (dissector_uses_key(flow_dissector,
1121 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1122 				key_tags = skb_flow_dissector_target(flow_dissector,
1123 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
1124 								     target_container);
1125 				key_tags->flow_label = ntohl(flow_label);
1126 			}
1127 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1128 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1129 				break;
1130 			}
1131 		}
1132 
1133 		__skb_flow_dissect_ipv6(skb, flow_dissector,
1134 					target_container, data, iph);
1135 
1136 		break;
1137 	}
1138 	case htons(ETH_P_8021AD):
1139 	case htons(ETH_P_8021Q): {
1140 		const struct vlan_hdr *vlan = NULL;
1141 		struct vlan_hdr _vlan;
1142 		__be16 saved_vlan_tpid = proto;
1143 
1144 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1145 		    skb && skb_vlan_tag_present(skb)) {
1146 			proto = skb->protocol;
1147 		} else {
1148 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1149 						    data, hlen, &_vlan);
1150 			if (!vlan) {
1151 				fdret = FLOW_DISSECT_RET_OUT_BAD;
1152 				break;
1153 			}
1154 
1155 			proto = vlan->h_vlan_encapsulated_proto;
1156 			nhoff += sizeof(*vlan);
1157 		}
1158 
1159 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1160 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1161 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1162 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1163 		} else {
1164 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1165 			break;
1166 		}
1167 
1168 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1169 			key_vlan = skb_flow_dissector_target(flow_dissector,
1170 							     dissector_vlan,
1171 							     target_container);
1172 
1173 			if (!vlan) {
1174 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1175 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1176 			} else {
1177 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1178 					VLAN_VID_MASK;
1179 				key_vlan->vlan_priority =
1180 					(ntohs(vlan->h_vlan_TCI) &
1181 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1182 			}
1183 			key_vlan->vlan_tpid = saved_vlan_tpid;
1184 		}
1185 
1186 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1187 		break;
1188 	}
1189 	case htons(ETH_P_PPP_SES): {
1190 		struct {
1191 			struct pppoe_hdr hdr;
1192 			__be16 proto;
1193 		} *hdr, _hdr;
1194 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1195 		if (!hdr) {
1196 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1197 			break;
1198 		}
1199 
1200 		nhoff += PPPOE_SES_HLEN;
1201 		switch (hdr->proto) {
1202 		case htons(PPP_IP):
1203 			proto = htons(ETH_P_IP);
1204 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1205 			break;
1206 		case htons(PPP_IPV6):
1207 			proto = htons(ETH_P_IPV6);
1208 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1209 			break;
1210 		default:
1211 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1212 			break;
1213 		}
1214 		break;
1215 	}
1216 	case htons(ETH_P_TIPC): {
1217 		struct tipc_basic_hdr *hdr, _hdr;
1218 
1219 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1220 					   data, hlen, &_hdr);
1221 		if (!hdr) {
1222 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1223 			break;
1224 		}
1225 
1226 		if (dissector_uses_key(flow_dissector,
1227 				       FLOW_DISSECTOR_KEY_TIPC)) {
1228 			key_addrs = skb_flow_dissector_target(flow_dissector,
1229 							      FLOW_DISSECTOR_KEY_TIPC,
1230 							      target_container);
1231 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1232 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1233 		}
1234 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1235 		break;
1236 	}
1237 
1238 	case htons(ETH_P_MPLS_UC):
1239 	case htons(ETH_P_MPLS_MC):
1240 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1241 						target_container, data,
1242 						nhoff, hlen, mpls_lse,
1243 						&mpls_el);
1244 		nhoff += sizeof(struct mpls_label);
1245 		mpls_lse++;
1246 		break;
1247 	case htons(ETH_P_FCOE):
1248 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1249 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1250 			break;
1251 		}
1252 
1253 		nhoff += FCOE_HEADER_LEN;
1254 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1255 		break;
1256 
1257 	case htons(ETH_P_ARP):
1258 	case htons(ETH_P_RARP):
1259 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1260 					       target_container, data,
1261 					       nhoff, hlen);
1262 		break;
1263 
1264 	case htons(ETH_P_BATMAN):
1265 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1266 						  &proto, &nhoff, hlen, flags);
1267 		break;
1268 
1269 	case htons(ETH_P_1588): {
1270 		struct ptp_header *hdr, _hdr;
1271 
1272 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1273 					   hlen, &_hdr);
1274 		if (!hdr) {
1275 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1276 			break;
1277 		}
1278 
1279 		nhoff += ntohs(hdr->message_length);
1280 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1281 		break;
1282 	}
1283 
1284 	default:
1285 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1286 		break;
1287 	}
1288 
1289 	/* Process result of proto processing */
1290 	switch (fdret) {
1291 	case FLOW_DISSECT_RET_OUT_GOOD:
1292 		goto out_good;
1293 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1294 		if (skb_flow_dissect_allowed(&num_hdrs))
1295 			goto proto_again;
1296 		goto out_good;
1297 	case FLOW_DISSECT_RET_CONTINUE:
1298 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1299 		break;
1300 	case FLOW_DISSECT_RET_OUT_BAD:
1301 	default:
1302 		goto out_bad;
1303 	}
1304 
1305 ip_proto_again:
1306 	fdret = FLOW_DISSECT_RET_CONTINUE;
1307 
1308 	switch (ip_proto) {
1309 	case IPPROTO_GRE:
1310 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1311 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1312 			break;
1313 		}
1314 
1315 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1316 					       target_container, data,
1317 					       &proto, &nhoff, &hlen, flags);
1318 		break;
1319 
1320 	case NEXTHDR_HOP:
1321 	case NEXTHDR_ROUTING:
1322 	case NEXTHDR_DEST: {
1323 		u8 _opthdr[2], *opthdr;
1324 
1325 		if (proto != htons(ETH_P_IPV6))
1326 			break;
1327 
1328 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1329 					      data, hlen, &_opthdr);
1330 		if (!opthdr) {
1331 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1332 			break;
1333 		}
1334 
1335 		ip_proto = opthdr[0];
1336 		nhoff += (opthdr[1] + 1) << 3;
1337 
1338 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1339 		break;
1340 	}
1341 	case NEXTHDR_FRAGMENT: {
1342 		struct frag_hdr _fh, *fh;
1343 
1344 		if (proto != htons(ETH_P_IPV6))
1345 			break;
1346 
1347 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1348 					  data, hlen, &_fh);
1349 
1350 		if (!fh) {
1351 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1352 			break;
1353 		}
1354 
1355 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1356 
1357 		nhoff += sizeof(_fh);
1358 		ip_proto = fh->nexthdr;
1359 
1360 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1361 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1362 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1363 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1364 				break;
1365 			}
1366 		}
1367 
1368 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1369 		break;
1370 	}
1371 	case IPPROTO_IPIP:
1372 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1373 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1374 			break;
1375 		}
1376 
1377 		proto = htons(ETH_P_IP);
1378 
1379 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1380 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1381 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1382 			break;
1383 		}
1384 
1385 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1386 		break;
1387 
1388 	case IPPROTO_IPV6:
1389 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1390 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1391 			break;
1392 		}
1393 
1394 		proto = htons(ETH_P_IPV6);
1395 
1396 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1397 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1398 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1399 			break;
1400 		}
1401 
1402 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1403 		break;
1404 
1405 
1406 	case IPPROTO_MPLS:
1407 		proto = htons(ETH_P_MPLS_UC);
1408 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1409 		break;
1410 
1411 	case IPPROTO_TCP:
1412 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1413 				       data, nhoff, hlen);
1414 		break;
1415 
1416 	case IPPROTO_ICMP:
1417 	case IPPROTO_ICMPV6:
1418 		__skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1419 					data, nhoff, hlen);
1420 		break;
1421 
1422 	default:
1423 		break;
1424 	}
1425 
1426 	if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1427 		__skb_flow_dissect_ports(skb, flow_dissector, target_container,
1428 					 data, nhoff, ip_proto, hlen);
1429 
1430 	/* Process result of IP proto processing */
1431 	switch (fdret) {
1432 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1433 		if (skb_flow_dissect_allowed(&num_hdrs))
1434 			goto proto_again;
1435 		break;
1436 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1437 		if (skb_flow_dissect_allowed(&num_hdrs))
1438 			goto ip_proto_again;
1439 		break;
1440 	case FLOW_DISSECT_RET_OUT_GOOD:
1441 	case FLOW_DISSECT_RET_CONTINUE:
1442 		break;
1443 	case FLOW_DISSECT_RET_OUT_BAD:
1444 	default:
1445 		goto out_bad;
1446 	}
1447 
1448 out_good:
1449 	ret = true;
1450 
1451 out:
1452 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1453 	key_basic->n_proto = proto;
1454 	key_basic->ip_proto = ip_proto;
1455 
1456 	return ret;
1457 
1458 out_bad:
1459 	ret = false;
1460 	goto out;
1461 }
1462 EXPORT_SYMBOL(__skb_flow_dissect);
1463 
1464 static siphash_key_t hashrnd __read_mostly;
__flow_hash_secret_init(void)1465 static __always_inline void __flow_hash_secret_init(void)
1466 {
1467 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1468 }
1469 
flow_keys_hash_start(const struct flow_keys * flow)1470 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1471 {
1472 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1473 	return &flow->FLOW_KEYS_HASH_START_FIELD;
1474 }
1475 
flow_keys_hash_length(const struct flow_keys * flow)1476 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1477 {
1478 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1479 
1480 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1481 
1482 	switch (flow->control.addr_type) {
1483 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1484 		diff -= sizeof(flow->addrs.v4addrs);
1485 		break;
1486 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1487 		diff -= sizeof(flow->addrs.v6addrs);
1488 		break;
1489 	case FLOW_DISSECTOR_KEY_TIPC:
1490 		diff -= sizeof(flow->addrs.tipckey);
1491 		break;
1492 	}
1493 	return sizeof(*flow) - diff;
1494 }
1495 
flow_get_u32_src(const struct flow_keys * flow)1496 __be32 flow_get_u32_src(const struct flow_keys *flow)
1497 {
1498 	switch (flow->control.addr_type) {
1499 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1500 		return flow->addrs.v4addrs.src;
1501 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1502 		return (__force __be32)ipv6_addr_hash(
1503 			&flow->addrs.v6addrs.src);
1504 	case FLOW_DISSECTOR_KEY_TIPC:
1505 		return flow->addrs.tipckey.key;
1506 	default:
1507 		return 0;
1508 	}
1509 }
1510 EXPORT_SYMBOL(flow_get_u32_src);
1511 
flow_get_u32_dst(const struct flow_keys * flow)1512 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1513 {
1514 	switch (flow->control.addr_type) {
1515 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1516 		return flow->addrs.v4addrs.dst;
1517 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1518 		return (__force __be32)ipv6_addr_hash(
1519 			&flow->addrs.v6addrs.dst);
1520 	default:
1521 		return 0;
1522 	}
1523 }
1524 EXPORT_SYMBOL(flow_get_u32_dst);
1525 
1526 /* Sort the source and destination IP and the ports,
1527  * to have consistent hash within the two directions
1528  */
__flow_hash_consistentify(struct flow_keys * keys)1529 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1530 {
1531 	int addr_diff, i;
1532 
1533 	switch (keys->control.addr_type) {
1534 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1535 		addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1536 			    (__force u32)keys->addrs.v4addrs.src;
1537 		if (addr_diff < 0)
1538 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1539 
1540 		if ((__force u16)keys->ports.dst <
1541 		    (__force u16)keys->ports.src) {
1542 			swap(keys->ports.src, keys->ports.dst);
1543 		}
1544 		break;
1545 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1546 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1547 				   &keys->addrs.v6addrs.src,
1548 				   sizeof(keys->addrs.v6addrs.dst));
1549 		if (addr_diff < 0) {
1550 			for (i = 0; i < 4; i++)
1551 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1552 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1553 		}
1554 		if ((__force u16)keys->ports.dst <
1555 		    (__force u16)keys->ports.src) {
1556 			swap(keys->ports.src, keys->ports.dst);
1557 		}
1558 		break;
1559 	}
1560 }
1561 
__flow_hash_from_keys(struct flow_keys * keys,const siphash_key_t * keyval)1562 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1563 					const siphash_key_t *keyval)
1564 {
1565 	u32 hash;
1566 
1567 	__flow_hash_consistentify(keys);
1568 
1569 	hash = siphash(flow_keys_hash_start(keys),
1570 		       flow_keys_hash_length(keys), keyval);
1571 	if (!hash)
1572 		hash = 1;
1573 
1574 	return hash;
1575 }
1576 
flow_hash_from_keys(struct flow_keys * keys)1577 u32 flow_hash_from_keys(struct flow_keys *keys)
1578 {
1579 	__flow_hash_secret_init();
1580 	return __flow_hash_from_keys(keys, &hashrnd);
1581 }
1582 EXPORT_SYMBOL(flow_hash_from_keys);
1583 
___skb_get_hash(const struct sk_buff * skb,struct flow_keys * keys,const siphash_key_t * keyval)1584 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1585 				  struct flow_keys *keys,
1586 				  const siphash_key_t *keyval)
1587 {
1588 	skb_flow_dissect_flow_keys(skb, keys,
1589 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1590 
1591 	return __flow_hash_from_keys(keys, keyval);
1592 }
1593 
1594 struct _flow_keys_digest_data {
1595 	__be16	n_proto;
1596 	u8	ip_proto;
1597 	u8	padding;
1598 	__be32	ports;
1599 	__be32	src;
1600 	__be32	dst;
1601 };
1602 
make_flow_keys_digest(struct flow_keys_digest * digest,const struct flow_keys * flow)1603 void make_flow_keys_digest(struct flow_keys_digest *digest,
1604 			   const struct flow_keys *flow)
1605 {
1606 	struct _flow_keys_digest_data *data =
1607 	    (struct _flow_keys_digest_data *)digest;
1608 
1609 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1610 
1611 	memset(digest, 0, sizeof(*digest));
1612 
1613 	data->n_proto = flow->basic.n_proto;
1614 	data->ip_proto = flow->basic.ip_proto;
1615 	data->ports = flow->ports.ports;
1616 	data->src = flow->addrs.v4addrs.src;
1617 	data->dst = flow->addrs.v4addrs.dst;
1618 }
1619 EXPORT_SYMBOL(make_flow_keys_digest);
1620 
1621 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1622 
__skb_get_hash_symmetric(const struct sk_buff * skb)1623 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1624 {
1625 	struct flow_keys keys;
1626 
1627 	__flow_hash_secret_init();
1628 
1629 	memset(&keys, 0, sizeof(keys));
1630 	__skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1631 			   &keys, NULL, 0, 0, 0,
1632 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1633 
1634 	return __flow_hash_from_keys(&keys, &hashrnd);
1635 }
1636 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1637 
1638 /**
1639  * __skb_get_hash: calculate a flow hash
1640  * @skb: sk_buff to calculate flow hash from
1641  *
1642  * This function calculates a flow hash based on src/dst addresses
1643  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1644  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1645  * if hash is a canonical 4-tuple hash over transport ports.
1646  */
__skb_get_hash(struct sk_buff * skb)1647 void __skb_get_hash(struct sk_buff *skb)
1648 {
1649 	struct flow_keys keys;
1650 	u32 hash;
1651 
1652 	__flow_hash_secret_init();
1653 
1654 	hash = ___skb_get_hash(skb, &keys, &hashrnd);
1655 
1656 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1657 }
1658 EXPORT_SYMBOL(__skb_get_hash);
1659 
skb_get_hash_perturb(const struct sk_buff * skb,const siphash_key_t * perturb)1660 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1661 			   const siphash_key_t *perturb)
1662 {
1663 	struct flow_keys keys;
1664 
1665 	return ___skb_get_hash(skb, &keys, perturb);
1666 }
1667 EXPORT_SYMBOL(skb_get_hash_perturb);
1668 
__skb_get_poff(const struct sk_buff * skb,const void * data,const struct flow_keys_basic * keys,int hlen)1669 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1670 		   const struct flow_keys_basic *keys, int hlen)
1671 {
1672 	u32 poff = keys->control.thoff;
1673 
1674 	/* skip L4 headers for fragments after the first */
1675 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1676 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1677 		return poff;
1678 
1679 	switch (keys->basic.ip_proto) {
1680 	case IPPROTO_TCP: {
1681 		/* access doff as u8 to avoid unaligned access */
1682 		const u8 *doff;
1683 		u8 _doff;
1684 
1685 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1686 					    data, hlen, &_doff);
1687 		if (!doff)
1688 			return poff;
1689 
1690 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1691 		break;
1692 	}
1693 	case IPPROTO_UDP:
1694 	case IPPROTO_UDPLITE:
1695 		poff += sizeof(struct udphdr);
1696 		break;
1697 	/* For the rest, we do not really care about header
1698 	 * extensions at this point for now.
1699 	 */
1700 	case IPPROTO_ICMP:
1701 		poff += sizeof(struct icmphdr);
1702 		break;
1703 	case IPPROTO_ICMPV6:
1704 		poff += sizeof(struct icmp6hdr);
1705 		break;
1706 	case IPPROTO_IGMP:
1707 		poff += sizeof(struct igmphdr);
1708 		break;
1709 	case IPPROTO_DCCP:
1710 		poff += sizeof(struct dccp_hdr);
1711 		break;
1712 	case IPPROTO_SCTP:
1713 		poff += sizeof(struct sctphdr);
1714 		break;
1715 	}
1716 
1717 	return poff;
1718 }
1719 
1720 /**
1721  * skb_get_poff - get the offset to the payload
1722  * @skb: sk_buff to get the payload offset from
1723  *
1724  * The function will get the offset to the payload as far as it could
1725  * be dissected.  The main user is currently BPF, so that we can dynamically
1726  * truncate packets without needing to push actual payload to the user
1727  * space and can analyze headers only, instead.
1728  */
skb_get_poff(const struct sk_buff * skb)1729 u32 skb_get_poff(const struct sk_buff *skb)
1730 {
1731 	struct flow_keys_basic keys;
1732 
1733 	if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1734 					      NULL, 0, 0, 0, 0))
1735 		return 0;
1736 
1737 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1738 }
1739 
__get_hash_from_flowi6(const struct flowi6 * fl6,struct flow_keys * keys)1740 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1741 {
1742 	memset(keys, 0, sizeof(*keys));
1743 
1744 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1745 	    sizeof(keys->addrs.v6addrs.src));
1746 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1747 	    sizeof(keys->addrs.v6addrs.dst));
1748 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1749 	keys->ports.src = fl6->fl6_sport;
1750 	keys->ports.dst = fl6->fl6_dport;
1751 	keys->keyid.keyid = fl6->fl6_gre_key;
1752 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1753 	keys->basic.ip_proto = fl6->flowi6_proto;
1754 
1755 	return flow_hash_from_keys(keys);
1756 }
1757 EXPORT_SYMBOL(__get_hash_from_flowi6);
1758 
1759 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1760 	{
1761 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1762 		.offset = offsetof(struct flow_keys, control),
1763 	},
1764 	{
1765 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1766 		.offset = offsetof(struct flow_keys, basic),
1767 	},
1768 	{
1769 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1770 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1771 	},
1772 	{
1773 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1774 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1775 	},
1776 	{
1777 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
1778 		.offset = offsetof(struct flow_keys, addrs.tipckey),
1779 	},
1780 	{
1781 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1782 		.offset = offsetof(struct flow_keys, ports),
1783 	},
1784 	{
1785 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
1786 		.offset = offsetof(struct flow_keys, vlan),
1787 	},
1788 	{
1789 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1790 		.offset = offsetof(struct flow_keys, tags),
1791 	},
1792 	{
1793 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1794 		.offset = offsetof(struct flow_keys, keyid),
1795 	},
1796 };
1797 
1798 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1799 	{
1800 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1801 		.offset = offsetof(struct flow_keys, control),
1802 	},
1803 	{
1804 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1805 		.offset = offsetof(struct flow_keys, basic),
1806 	},
1807 	{
1808 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1809 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
1810 	},
1811 	{
1812 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1813 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
1814 	},
1815 	{
1816 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
1817 		.offset = offsetof(struct flow_keys, ports),
1818 	},
1819 };
1820 
1821 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1822 	{
1823 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
1824 		.offset = offsetof(struct flow_keys, control),
1825 	},
1826 	{
1827 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
1828 		.offset = offsetof(struct flow_keys, basic),
1829 	},
1830 };
1831 
1832 struct flow_dissector flow_keys_dissector __read_mostly;
1833 EXPORT_SYMBOL(flow_keys_dissector);
1834 
1835 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1836 EXPORT_SYMBOL(flow_keys_basic_dissector);
1837 
init_default_flow_dissectors(void)1838 static int __init init_default_flow_dissectors(void)
1839 {
1840 	skb_flow_dissector_init(&flow_keys_dissector,
1841 				flow_keys_dissector_keys,
1842 				ARRAY_SIZE(flow_keys_dissector_keys));
1843 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1844 				flow_keys_dissector_symmetric_keys,
1845 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1846 	skb_flow_dissector_init(&flow_keys_basic_dissector,
1847 				flow_keys_basic_dissector_keys,
1848 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
1849 	return 0;
1850 }
1851 core_initcall(init_default_flow_dissectors);
1852