xref: /linux/net/core/flow_dissector.c (revision a1ff5a7d78a036d6c2178ee5acd6ba4946243800)
1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/kernel.h>
3 #include <linux/skbuff.h>
4 #include <linux/export.h>
5 #include <linux/ip.h>
6 #include <linux/ipv6.h>
7 #include <linux/if_vlan.h>
8 #include <linux/filter.h>
9 #include <net/dsa.h>
10 #include <net/dst_metadata.h>
11 #include <net/ip.h>
12 #include <net/ipv6.h>
13 #include <net/gre.h>
14 #include <net/pptp.h>
15 #include <net/tipc.h>
16 #include <linux/igmp.h>
17 #include <linux/icmp.h>
18 #include <linux/sctp.h>
19 #include <linux/dccp.h>
20 #include <linux/if_tunnel.h>
21 #include <linux/if_pppox.h>
22 #include <linux/ppp_defs.h>
23 #include <linux/stddef.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_hsr.h>
26 #include <linux/mpls.h>
27 #include <linux/tcp.h>
28 #include <linux/ptp_classify.h>
29 #include <net/flow_dissector.h>
30 #include <net/pkt_cls.h>
31 #include <scsi/fc/fc_fcoe.h>
32 #include <uapi/linux/batadv_packet.h>
33 #include <linux/bpf.h>
34 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
35 #include <net/netfilter/nf_conntrack_core.h>
36 #include <net/netfilter/nf_conntrack_labels.h>
37 #endif
38 #include <linux/bpf-netns.h>
39 
dissector_set_key(struct flow_dissector * flow_dissector,enum flow_dissector_key_id key_id)40 static void dissector_set_key(struct flow_dissector *flow_dissector,
41 			      enum flow_dissector_key_id key_id)
42 {
43 	flow_dissector->used_keys |= (1ULL << key_id);
44 }
45 
skb_flow_dissector_init(struct flow_dissector * flow_dissector,const struct flow_dissector_key * key,unsigned int key_count)46 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
47 			     const struct flow_dissector_key *key,
48 			     unsigned int key_count)
49 {
50 	unsigned int i;
51 
52 	memset(flow_dissector, 0, sizeof(*flow_dissector));
53 
54 	for (i = 0; i < key_count; i++, key++) {
55 		/* User should make sure that every key target offset is within
56 		 * boundaries of unsigned short.
57 		 */
58 		BUG_ON(key->offset > USHRT_MAX);
59 		BUG_ON(dissector_uses_key(flow_dissector,
60 					  key->key_id));
61 
62 		dissector_set_key(flow_dissector, key->key_id);
63 		flow_dissector->offset[key->key_id] = key->offset;
64 	}
65 
66 	/* Ensure that the dissector always includes control and basic key.
67 	 * That way we are able to avoid handling lack of these in fast path.
68 	 */
69 	BUG_ON(!dissector_uses_key(flow_dissector,
70 				   FLOW_DISSECTOR_KEY_CONTROL));
71 	BUG_ON(!dissector_uses_key(flow_dissector,
72 				   FLOW_DISSECTOR_KEY_BASIC));
73 }
74 EXPORT_SYMBOL(skb_flow_dissector_init);
75 
76 #ifdef CONFIG_BPF_SYSCALL
flow_dissector_bpf_prog_attach_check(struct net * net,struct bpf_prog * prog)77 int flow_dissector_bpf_prog_attach_check(struct net *net,
78 					 struct bpf_prog *prog)
79 {
80 	enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
81 
82 	if (net == &init_net) {
83 		/* BPF flow dissector in the root namespace overrides
84 		 * any per-net-namespace one. When attaching to root,
85 		 * make sure we don't have any BPF program attached
86 		 * to the non-root namespaces.
87 		 */
88 		struct net *ns;
89 
90 		for_each_net(ns) {
91 			if (ns == &init_net)
92 				continue;
93 			if (rcu_access_pointer(ns->bpf.run_array[type]))
94 				return -EEXIST;
95 		}
96 	} else {
97 		/* Make sure root flow dissector is not attached
98 		 * when attaching to the non-root namespace.
99 		 */
100 		if (rcu_access_pointer(init_net.bpf.run_array[type]))
101 			return -EEXIST;
102 	}
103 
104 	return 0;
105 }
106 #endif /* CONFIG_BPF_SYSCALL */
107 
108 /**
109  * __skb_flow_get_ports - extract the upper layer ports and return them
110  * @skb: sk_buff to extract the ports from
111  * @thoff: transport header offset
112  * @ip_proto: protocol for which to get port offset
113  * @data: raw buffer pointer to the packet, if NULL use skb->data
114  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
115  *
116  * The function will try to retrieve the ports at offset thoff + poff where poff
117  * is the protocol port offset returned from proto_ports_offset
118  */
__skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto,const void * data,int hlen)119 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
120 			    const void *data, int hlen)
121 {
122 	int poff = proto_ports_offset(ip_proto);
123 
124 	if (!data) {
125 		data = skb->data;
126 		hlen = skb_headlen(skb);
127 	}
128 
129 	if (poff >= 0) {
130 		__be32 *ports, _ports;
131 
132 		ports = __skb_header_pointer(skb, thoff + poff,
133 					     sizeof(_ports), data, hlen, &_ports);
134 		if (ports)
135 			return *ports;
136 	}
137 
138 	return 0;
139 }
140 EXPORT_SYMBOL(__skb_flow_get_ports);
141 
icmp_has_id(u8 type)142 static bool icmp_has_id(u8 type)
143 {
144 	switch (type) {
145 	case ICMP_ECHO:
146 	case ICMP_ECHOREPLY:
147 	case ICMP_TIMESTAMP:
148 	case ICMP_TIMESTAMPREPLY:
149 	case ICMPV6_ECHO_REQUEST:
150 	case ICMPV6_ECHO_REPLY:
151 		return true;
152 	}
153 
154 	return false;
155 }
156 
157 /**
158  * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
159  * @skb: sk_buff to extract from
160  * @key_icmp: struct flow_dissector_key_icmp to fill
161  * @data: raw buffer pointer to the packet
162  * @thoff: offset to extract at
163  * @hlen: packet header length
164  */
skb_flow_get_icmp_tci(const struct sk_buff * skb,struct flow_dissector_key_icmp * key_icmp,const void * data,int thoff,int hlen)165 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
166 			   struct flow_dissector_key_icmp *key_icmp,
167 			   const void *data, int thoff, int hlen)
168 {
169 	struct icmphdr *ih, _ih;
170 
171 	ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
172 	if (!ih)
173 		return;
174 
175 	key_icmp->type = ih->type;
176 	key_icmp->code = ih->code;
177 
178 	/* As we use 0 to signal that the Id field is not present,
179 	 * avoid confusion with packets without such field
180 	 */
181 	if (icmp_has_id(ih->type))
182 		key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
183 	else
184 		key_icmp->id = 0;
185 }
186 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
187 
188 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
189  * using skb_flow_get_icmp_tci().
190  */
__skb_flow_dissect_icmp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int thoff,int hlen)191 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
192 				    struct flow_dissector *flow_dissector,
193 				    void *target_container, const void *data,
194 				    int thoff, int hlen)
195 {
196 	struct flow_dissector_key_icmp *key_icmp;
197 
198 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
199 		return;
200 
201 	key_icmp = skb_flow_dissector_target(flow_dissector,
202 					     FLOW_DISSECTOR_KEY_ICMP,
203 					     target_container);
204 
205 	skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
206 }
207 
__skb_flow_dissect_ah(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)208 static void __skb_flow_dissect_ah(const struct sk_buff *skb,
209 				  struct flow_dissector *flow_dissector,
210 				  void *target_container, const void *data,
211 				  int nhoff, int hlen)
212 {
213 	struct flow_dissector_key_ipsec *key_ah;
214 	struct ip_auth_hdr _hdr, *hdr;
215 
216 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
217 		return;
218 
219 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
220 	if (!hdr)
221 		return;
222 
223 	key_ah = skb_flow_dissector_target(flow_dissector,
224 					   FLOW_DISSECTOR_KEY_IPSEC,
225 					   target_container);
226 
227 	key_ah->spi = hdr->spi;
228 }
229 
__skb_flow_dissect_esp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)230 static void __skb_flow_dissect_esp(const struct sk_buff *skb,
231 				   struct flow_dissector *flow_dissector,
232 				   void *target_container, const void *data,
233 				   int nhoff, int hlen)
234 {
235 	struct flow_dissector_key_ipsec *key_esp;
236 	struct ip_esp_hdr _hdr, *hdr;
237 
238 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
239 		return;
240 
241 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
242 	if (!hdr)
243 		return;
244 
245 	key_esp = skb_flow_dissector_target(flow_dissector,
246 					    FLOW_DISSECTOR_KEY_IPSEC,
247 					    target_container);
248 
249 	key_esp->spi = hdr->spi;
250 }
251 
__skb_flow_dissect_l2tpv3(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)252 static void __skb_flow_dissect_l2tpv3(const struct sk_buff *skb,
253 				      struct flow_dissector *flow_dissector,
254 				      void *target_container, const void *data,
255 				      int nhoff, int hlen)
256 {
257 	struct flow_dissector_key_l2tpv3 *key_l2tpv3;
258 	struct {
259 		__be32 session_id;
260 	} *hdr, _hdr;
261 
262 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3))
263 		return;
264 
265 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
266 	if (!hdr)
267 		return;
268 
269 	key_l2tpv3 = skb_flow_dissector_target(flow_dissector,
270 					       FLOW_DISSECTOR_KEY_L2TPV3,
271 					       target_container);
272 
273 	key_l2tpv3->session_id = hdr->session_id;
274 }
275 
skb_flow_dissect_meta(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)276 void skb_flow_dissect_meta(const struct sk_buff *skb,
277 			   struct flow_dissector *flow_dissector,
278 			   void *target_container)
279 {
280 	struct flow_dissector_key_meta *meta;
281 
282 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
283 		return;
284 
285 	meta = skb_flow_dissector_target(flow_dissector,
286 					 FLOW_DISSECTOR_KEY_META,
287 					 target_container);
288 	meta->ingress_ifindex = skb->skb_iif;
289 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
290 	if (tc_skb_ext_tc_enabled()) {
291 		struct tc_skb_ext *ext;
292 
293 		ext = skb_ext_find(skb, TC_SKB_EXT);
294 		if (ext)
295 			meta->l2_miss = ext->l2_miss;
296 	}
297 #endif
298 }
299 EXPORT_SYMBOL(skb_flow_dissect_meta);
300 
301 static void
skb_flow_dissect_set_enc_control(enum flow_dissector_key_id type,u32 ctrl_flags,struct flow_dissector * flow_dissector,void * target_container)302 skb_flow_dissect_set_enc_control(enum flow_dissector_key_id type,
303 				 u32 ctrl_flags,
304 				 struct flow_dissector *flow_dissector,
305 				 void *target_container)
306 {
307 	struct flow_dissector_key_control *ctrl;
308 
309 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
310 		return;
311 
312 	ctrl = skb_flow_dissector_target(flow_dissector,
313 					 FLOW_DISSECTOR_KEY_ENC_CONTROL,
314 					 target_container);
315 	ctrl->addr_type = type;
316 	ctrl->flags = ctrl_flags;
317 }
318 
319 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)320 skb_flow_dissect_ct(const struct sk_buff *skb,
321 		    struct flow_dissector *flow_dissector,
322 		    void *target_container, u16 *ctinfo_map,
323 		    size_t mapsize, bool post_ct, u16 zone)
324 {
325 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
326 	struct flow_dissector_key_ct *key;
327 	enum ip_conntrack_info ctinfo;
328 	struct nf_conn_labels *cl;
329 	struct nf_conn *ct;
330 
331 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
332 		return;
333 
334 	ct = nf_ct_get(skb, &ctinfo);
335 	if (!ct && !post_ct)
336 		return;
337 
338 	key = skb_flow_dissector_target(flow_dissector,
339 					FLOW_DISSECTOR_KEY_CT,
340 					target_container);
341 
342 	if (!ct) {
343 		key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
344 				TCA_FLOWER_KEY_CT_FLAGS_INVALID;
345 		key->ct_zone = zone;
346 		return;
347 	}
348 
349 	if (ctinfo < mapsize)
350 		key->ct_state = ctinfo_map[ctinfo];
351 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
352 	key->ct_zone = ct->zone.id;
353 #endif
354 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
355 	key->ct_mark = READ_ONCE(ct->mark);
356 #endif
357 
358 	cl = nf_ct_labels_find(ct);
359 	if (cl)
360 		memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
361 #endif /* CONFIG_NF_CONNTRACK */
362 }
363 EXPORT_SYMBOL(skb_flow_dissect_ct);
364 
365 void
skb_flow_dissect_tunnel_info(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)366 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
367 			     struct flow_dissector *flow_dissector,
368 			     void *target_container)
369 {
370 	struct ip_tunnel_info *info;
371 	struct ip_tunnel_key *key;
372 	u32 ctrl_flags = 0;
373 
374 	/* A quick check to see if there might be something to do. */
375 	if (!dissector_uses_key(flow_dissector,
376 				FLOW_DISSECTOR_KEY_ENC_KEYID) &&
377 	    !dissector_uses_key(flow_dissector,
378 				FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
379 	    !dissector_uses_key(flow_dissector,
380 				FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
381 	    !dissector_uses_key(flow_dissector,
382 				FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
383 	    !dissector_uses_key(flow_dissector,
384 				FLOW_DISSECTOR_KEY_ENC_PORTS) &&
385 	    !dissector_uses_key(flow_dissector,
386 				FLOW_DISSECTOR_KEY_ENC_IP) &&
387 	    !dissector_uses_key(flow_dissector,
388 				FLOW_DISSECTOR_KEY_ENC_OPTS))
389 		return;
390 
391 	info = skb_tunnel_info(skb);
392 	if (!info)
393 		return;
394 
395 	key = &info->key;
396 
397 	if (test_bit(IP_TUNNEL_CSUM_BIT, key->tun_flags))
398 		ctrl_flags |= FLOW_DIS_F_TUNNEL_CSUM;
399 	if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, key->tun_flags))
400 		ctrl_flags |= FLOW_DIS_F_TUNNEL_DONT_FRAGMENT;
401 	if (test_bit(IP_TUNNEL_OAM_BIT, key->tun_flags))
402 		ctrl_flags |= FLOW_DIS_F_TUNNEL_OAM;
403 	if (test_bit(IP_TUNNEL_CRIT_OPT_BIT, key->tun_flags))
404 		ctrl_flags |= FLOW_DIS_F_TUNNEL_CRIT_OPT;
405 
406 	switch (ip_tunnel_info_af(info)) {
407 	case AF_INET:
408 		skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
409 						 ctrl_flags, flow_dissector,
410 						 target_container);
411 		if (dissector_uses_key(flow_dissector,
412 				       FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
413 			struct flow_dissector_key_ipv4_addrs *ipv4;
414 
415 			ipv4 = skb_flow_dissector_target(flow_dissector,
416 							 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
417 							 target_container);
418 			ipv4->src = key->u.ipv4.src;
419 			ipv4->dst = key->u.ipv4.dst;
420 		}
421 		break;
422 	case AF_INET6:
423 		skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
424 						 ctrl_flags, flow_dissector,
425 						 target_container);
426 		if (dissector_uses_key(flow_dissector,
427 				       FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
428 			struct flow_dissector_key_ipv6_addrs *ipv6;
429 
430 			ipv6 = skb_flow_dissector_target(flow_dissector,
431 							 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
432 							 target_container);
433 			ipv6->src = key->u.ipv6.src;
434 			ipv6->dst = key->u.ipv6.dst;
435 		}
436 		break;
437 	default:
438 		skb_flow_dissect_set_enc_control(0, ctrl_flags, flow_dissector,
439 						 target_container);
440 		break;
441 	}
442 
443 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
444 		struct flow_dissector_key_keyid *keyid;
445 
446 		keyid = skb_flow_dissector_target(flow_dissector,
447 						  FLOW_DISSECTOR_KEY_ENC_KEYID,
448 						  target_container);
449 		keyid->keyid = tunnel_id_to_key32(key->tun_id);
450 	}
451 
452 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
453 		struct flow_dissector_key_ports *tp;
454 
455 		tp = skb_flow_dissector_target(flow_dissector,
456 					       FLOW_DISSECTOR_KEY_ENC_PORTS,
457 					       target_container);
458 		tp->src = key->tp_src;
459 		tp->dst = key->tp_dst;
460 	}
461 
462 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
463 		struct flow_dissector_key_ip *ip;
464 
465 		ip = skb_flow_dissector_target(flow_dissector,
466 					       FLOW_DISSECTOR_KEY_ENC_IP,
467 					       target_container);
468 		ip->tos = key->tos;
469 		ip->ttl = key->ttl;
470 	}
471 
472 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
473 		struct flow_dissector_key_enc_opts *enc_opt;
474 		IP_TUNNEL_DECLARE_FLAGS(flags) = { };
475 		u32 val;
476 
477 		enc_opt = skb_flow_dissector_target(flow_dissector,
478 						    FLOW_DISSECTOR_KEY_ENC_OPTS,
479 						    target_container);
480 
481 		if (!info->options_len)
482 			return;
483 
484 		enc_opt->len = info->options_len;
485 		ip_tunnel_info_opts_get(enc_opt->data, info);
486 
487 		ip_tunnel_set_options_present(flags);
488 		ip_tunnel_flags_and(flags, info->key.tun_flags, flags);
489 
490 		val = find_next_bit(flags, __IP_TUNNEL_FLAG_NUM,
491 				    IP_TUNNEL_GENEVE_OPT_BIT);
492 		enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val : 0;
493 	}
494 }
495 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
496 
skb_flow_dissect_hash(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)497 void skb_flow_dissect_hash(const struct sk_buff *skb,
498 			   struct flow_dissector *flow_dissector,
499 			   void *target_container)
500 {
501 	struct flow_dissector_key_hash *key;
502 
503 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
504 		return;
505 
506 	key = skb_flow_dissector_target(flow_dissector,
507 					FLOW_DISSECTOR_KEY_HASH,
508 					target_container);
509 
510 	key->hash = skb_get_hash_raw(skb);
511 }
512 EXPORT_SYMBOL(skb_flow_dissect_hash);
513 
514 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)515 __skb_flow_dissect_mpls(const struct sk_buff *skb,
516 			struct flow_dissector *flow_dissector,
517 			void *target_container, const void *data, int nhoff,
518 			int hlen, int lse_index, bool *entropy_label)
519 {
520 	struct mpls_label *hdr, _hdr;
521 	u32 entry, label, bos;
522 
523 	if (!dissector_uses_key(flow_dissector,
524 				FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
525 	    !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
526 		return FLOW_DISSECT_RET_OUT_GOOD;
527 
528 	if (lse_index >= FLOW_DIS_MPLS_MAX)
529 		return FLOW_DISSECT_RET_OUT_GOOD;
530 
531 	hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
532 				   hlen, &_hdr);
533 	if (!hdr)
534 		return FLOW_DISSECT_RET_OUT_BAD;
535 
536 	entry = ntohl(hdr->entry);
537 	label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
538 	bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
539 
540 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
541 		struct flow_dissector_key_mpls *key_mpls;
542 		struct flow_dissector_mpls_lse *lse;
543 
544 		key_mpls = skb_flow_dissector_target(flow_dissector,
545 						     FLOW_DISSECTOR_KEY_MPLS,
546 						     target_container);
547 		lse = &key_mpls->ls[lse_index];
548 
549 		lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
550 		lse->mpls_bos = bos;
551 		lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
552 		lse->mpls_label = label;
553 		dissector_set_mpls_lse(key_mpls, lse_index);
554 	}
555 
556 	if (*entropy_label &&
557 	    dissector_uses_key(flow_dissector,
558 			       FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
559 		struct flow_dissector_key_keyid *key_keyid;
560 
561 		key_keyid = skb_flow_dissector_target(flow_dissector,
562 						      FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
563 						      target_container);
564 		key_keyid->keyid = cpu_to_be32(label);
565 	}
566 
567 	*entropy_label = label == MPLS_LABEL_ENTROPY;
568 
569 	return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
570 }
571 
572 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)573 __skb_flow_dissect_arp(const struct sk_buff *skb,
574 		       struct flow_dissector *flow_dissector,
575 		       void *target_container, const void *data,
576 		       int nhoff, int hlen)
577 {
578 	struct flow_dissector_key_arp *key_arp;
579 	struct {
580 		unsigned char ar_sha[ETH_ALEN];
581 		unsigned char ar_sip[4];
582 		unsigned char ar_tha[ETH_ALEN];
583 		unsigned char ar_tip[4];
584 	} *arp_eth, _arp_eth;
585 	const struct arphdr *arp;
586 	struct arphdr _arp;
587 
588 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
589 		return FLOW_DISSECT_RET_OUT_GOOD;
590 
591 	arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
592 				   hlen, &_arp);
593 	if (!arp)
594 		return FLOW_DISSECT_RET_OUT_BAD;
595 
596 	if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
597 	    arp->ar_pro != htons(ETH_P_IP) ||
598 	    arp->ar_hln != ETH_ALEN ||
599 	    arp->ar_pln != 4 ||
600 	    (arp->ar_op != htons(ARPOP_REPLY) &&
601 	     arp->ar_op != htons(ARPOP_REQUEST)))
602 		return FLOW_DISSECT_RET_OUT_BAD;
603 
604 	arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
605 				       sizeof(_arp_eth), data,
606 				       hlen, &_arp_eth);
607 	if (!arp_eth)
608 		return FLOW_DISSECT_RET_OUT_BAD;
609 
610 	key_arp = skb_flow_dissector_target(flow_dissector,
611 					    FLOW_DISSECTOR_KEY_ARP,
612 					    target_container);
613 
614 	memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
615 	memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
616 
617 	/* Only store the lower byte of the opcode;
618 	 * this covers ARPOP_REPLY and ARPOP_REQUEST.
619 	 */
620 	key_arp->op = ntohs(arp->ar_op) & 0xff;
621 
622 	ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
623 	ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
624 
625 	return FLOW_DISSECT_RET_OUT_GOOD;
626 }
627 
628 static enum flow_dissect_ret
__skb_flow_dissect_cfm(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)629 __skb_flow_dissect_cfm(const struct sk_buff *skb,
630 		       struct flow_dissector *flow_dissector,
631 		       void *target_container, const void *data,
632 		       int nhoff, int hlen)
633 {
634 	struct flow_dissector_key_cfm *key, *hdr, _hdr;
635 
636 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM))
637 		return FLOW_DISSECT_RET_OUT_GOOD;
638 
639 	hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr);
640 	if (!hdr)
641 		return FLOW_DISSECT_RET_OUT_BAD;
642 
643 	key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM,
644 					target_container);
645 
646 	key->mdl_ver = hdr->mdl_ver;
647 	key->opcode = hdr->opcode;
648 
649 	return FLOW_DISSECT_RET_OUT_GOOD;
650 }
651 
652 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)653 __skb_flow_dissect_gre(const struct sk_buff *skb,
654 		       struct flow_dissector_key_control *key_control,
655 		       struct flow_dissector *flow_dissector,
656 		       void *target_container, const void *data,
657 		       __be16 *p_proto, int *p_nhoff, int *p_hlen,
658 		       unsigned int flags)
659 {
660 	struct flow_dissector_key_keyid *key_keyid;
661 	struct gre_base_hdr *hdr, _hdr;
662 	int offset = 0;
663 	u16 gre_ver;
664 
665 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
666 				   data, *p_hlen, &_hdr);
667 	if (!hdr)
668 		return FLOW_DISSECT_RET_OUT_BAD;
669 
670 	/* Only look inside GRE without routing */
671 	if (hdr->flags & GRE_ROUTING)
672 		return FLOW_DISSECT_RET_OUT_GOOD;
673 
674 	/* Only look inside GRE for version 0 and 1 */
675 	gre_ver = ntohs(hdr->flags & GRE_VERSION);
676 	if (gre_ver > 1)
677 		return FLOW_DISSECT_RET_OUT_GOOD;
678 
679 	*p_proto = hdr->protocol;
680 	if (gre_ver) {
681 		/* Version1 must be PPTP, and check the flags */
682 		if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
683 			return FLOW_DISSECT_RET_OUT_GOOD;
684 	}
685 
686 	offset += sizeof(struct gre_base_hdr);
687 
688 	if (hdr->flags & GRE_CSUM)
689 		offset += sizeof_field(struct gre_full_hdr, csum) +
690 			  sizeof_field(struct gre_full_hdr, reserved1);
691 
692 	if (hdr->flags & GRE_KEY) {
693 		const __be32 *keyid;
694 		__be32 _keyid;
695 
696 		keyid = __skb_header_pointer(skb, *p_nhoff + offset,
697 					     sizeof(_keyid),
698 					     data, *p_hlen, &_keyid);
699 		if (!keyid)
700 			return FLOW_DISSECT_RET_OUT_BAD;
701 
702 		if (dissector_uses_key(flow_dissector,
703 				       FLOW_DISSECTOR_KEY_GRE_KEYID)) {
704 			key_keyid = skb_flow_dissector_target(flow_dissector,
705 							      FLOW_DISSECTOR_KEY_GRE_KEYID,
706 							      target_container);
707 			if (gre_ver == 0)
708 				key_keyid->keyid = *keyid;
709 			else
710 				key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
711 		}
712 		offset += sizeof_field(struct gre_full_hdr, key);
713 	}
714 
715 	if (hdr->flags & GRE_SEQ)
716 		offset += sizeof_field(struct pptp_gre_header, seq);
717 
718 	if (gre_ver == 0) {
719 		if (*p_proto == htons(ETH_P_TEB)) {
720 			const struct ethhdr *eth;
721 			struct ethhdr _eth;
722 
723 			eth = __skb_header_pointer(skb, *p_nhoff + offset,
724 						   sizeof(_eth),
725 						   data, *p_hlen, &_eth);
726 			if (!eth)
727 				return FLOW_DISSECT_RET_OUT_BAD;
728 			*p_proto = eth->h_proto;
729 			offset += sizeof(*eth);
730 
731 			/* Cap headers that we access via pointers at the
732 			 * end of the Ethernet header as our maximum alignment
733 			 * at that point is only 2 bytes.
734 			 */
735 			if (NET_IP_ALIGN)
736 				*p_hlen = *p_nhoff + offset;
737 		}
738 	} else { /* version 1, must be PPTP */
739 		u8 _ppp_hdr[PPP_HDRLEN];
740 		u8 *ppp_hdr;
741 
742 		if (hdr->flags & GRE_ACK)
743 			offset += sizeof_field(struct pptp_gre_header, ack);
744 
745 		ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
746 					       sizeof(_ppp_hdr),
747 					       data, *p_hlen, _ppp_hdr);
748 		if (!ppp_hdr)
749 			return FLOW_DISSECT_RET_OUT_BAD;
750 
751 		switch (PPP_PROTOCOL(ppp_hdr)) {
752 		case PPP_IP:
753 			*p_proto = htons(ETH_P_IP);
754 			break;
755 		case PPP_IPV6:
756 			*p_proto = htons(ETH_P_IPV6);
757 			break;
758 		default:
759 			/* Could probably catch some more like MPLS */
760 			break;
761 		}
762 
763 		offset += PPP_HDRLEN;
764 	}
765 
766 	*p_nhoff += offset;
767 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
768 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
769 		return FLOW_DISSECT_RET_OUT_GOOD;
770 
771 	return FLOW_DISSECT_RET_PROTO_AGAIN;
772 }
773 
774 /**
775  * __skb_flow_dissect_batadv() - dissect batman-adv header
776  * @skb: sk_buff to with the batman-adv header
777  * @key_control: flow dissectors control key
778  * @data: raw buffer pointer to the packet, if NULL use skb->data
779  * @p_proto: pointer used to update the protocol to process next
780  * @p_nhoff: pointer used to update inner network header offset
781  * @hlen: packet header length
782  * @flags: any combination of FLOW_DISSECTOR_F_*
783  *
784  * ETH_P_BATMAN packets are tried to be dissected. Only
785  * &struct batadv_unicast packets are actually processed because they contain an
786  * inner ethernet header and are usually followed by actual network header. This
787  * allows the flow dissector to continue processing the packet.
788  *
789  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
790  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
791  *  otherwise FLOW_DISSECT_RET_OUT_BAD
792  */
793 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)794 __skb_flow_dissect_batadv(const struct sk_buff *skb,
795 			  struct flow_dissector_key_control *key_control,
796 			  const void *data, __be16 *p_proto, int *p_nhoff,
797 			  int hlen, unsigned int flags)
798 {
799 	struct {
800 		struct batadv_unicast_packet batadv_unicast;
801 		struct ethhdr eth;
802 	} *hdr, _hdr;
803 
804 	hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
805 				   &_hdr);
806 	if (!hdr)
807 		return FLOW_DISSECT_RET_OUT_BAD;
808 
809 	if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
810 		return FLOW_DISSECT_RET_OUT_BAD;
811 
812 	if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
813 		return FLOW_DISSECT_RET_OUT_BAD;
814 
815 	*p_proto = hdr->eth.h_proto;
816 	*p_nhoff += sizeof(*hdr);
817 
818 	key_control->flags |= FLOW_DIS_ENCAPSULATION;
819 	if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
820 		return FLOW_DISSECT_RET_OUT_GOOD;
821 
822 	return FLOW_DISSECT_RET_PROTO_AGAIN;
823 }
824 
825 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)826 __skb_flow_dissect_tcp(const struct sk_buff *skb,
827 		       struct flow_dissector *flow_dissector,
828 		       void *target_container, const void *data,
829 		       int thoff, int hlen)
830 {
831 	struct flow_dissector_key_tcp *key_tcp;
832 	struct tcphdr *th, _th;
833 
834 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
835 		return;
836 
837 	th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
838 	if (!th)
839 		return;
840 
841 	if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
842 		return;
843 
844 	key_tcp = skb_flow_dissector_target(flow_dissector,
845 					    FLOW_DISSECTOR_KEY_TCP,
846 					    target_container);
847 	key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
848 }
849 
850 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)851 __skb_flow_dissect_ports(const struct sk_buff *skb,
852 			 struct flow_dissector *flow_dissector,
853 			 void *target_container, const void *data,
854 			 int nhoff, u8 ip_proto, int hlen)
855 {
856 	enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
857 	struct flow_dissector_key_ports *key_ports;
858 
859 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
860 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
861 	else if (dissector_uses_key(flow_dissector,
862 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
863 		dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
864 
865 	if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
866 		return;
867 
868 	key_ports = skb_flow_dissector_target(flow_dissector,
869 					      dissector_ports,
870 					      target_container);
871 	key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
872 						data, hlen);
873 }
874 
875 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)876 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
877 			struct flow_dissector *flow_dissector,
878 			void *target_container, const void *data,
879 			const struct iphdr *iph)
880 {
881 	struct flow_dissector_key_ip *key_ip;
882 
883 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
884 		return;
885 
886 	key_ip = skb_flow_dissector_target(flow_dissector,
887 					   FLOW_DISSECTOR_KEY_IP,
888 					   target_container);
889 	key_ip->tos = iph->tos;
890 	key_ip->ttl = iph->ttl;
891 }
892 
893 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)894 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
895 			struct flow_dissector *flow_dissector,
896 			void *target_container, const void *data,
897 			const struct ipv6hdr *iph)
898 {
899 	struct flow_dissector_key_ip *key_ip;
900 
901 	if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
902 		return;
903 
904 	key_ip = skb_flow_dissector_target(flow_dissector,
905 					   FLOW_DISSECTOR_KEY_IP,
906 					   target_container);
907 	key_ip->tos = ipv6_get_dsfield(iph);
908 	key_ip->ttl = iph->hop_limit;
909 }
910 
911 /* Maximum number of protocol headers that can be parsed in
912  * __skb_flow_dissect
913  */
914 #define MAX_FLOW_DISSECT_HDRS	15
915 
skb_flow_dissect_allowed(int * num_hdrs)916 static bool skb_flow_dissect_allowed(int *num_hdrs)
917 {
918 	++*num_hdrs;
919 
920 	return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
921 }
922 
__skb_flow_bpf_to_target(const struct bpf_flow_keys * flow_keys,struct flow_dissector * flow_dissector,void * target_container)923 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
924 				     struct flow_dissector *flow_dissector,
925 				     void *target_container)
926 {
927 	struct flow_dissector_key_ports *key_ports = NULL;
928 	struct flow_dissector_key_control *key_control;
929 	struct flow_dissector_key_basic *key_basic;
930 	struct flow_dissector_key_addrs *key_addrs;
931 	struct flow_dissector_key_tags *key_tags;
932 
933 	key_control = skb_flow_dissector_target(flow_dissector,
934 						FLOW_DISSECTOR_KEY_CONTROL,
935 						target_container);
936 	key_control->thoff = flow_keys->thoff;
937 	if (flow_keys->is_frag)
938 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
939 	if (flow_keys->is_first_frag)
940 		key_control->flags |= FLOW_DIS_FIRST_FRAG;
941 	if (flow_keys->is_encap)
942 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
943 
944 	key_basic = skb_flow_dissector_target(flow_dissector,
945 					      FLOW_DISSECTOR_KEY_BASIC,
946 					      target_container);
947 	key_basic->n_proto = flow_keys->n_proto;
948 	key_basic->ip_proto = flow_keys->ip_proto;
949 
950 	if (flow_keys->addr_proto == ETH_P_IP &&
951 	    dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
952 		key_addrs = skb_flow_dissector_target(flow_dissector,
953 						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
954 						      target_container);
955 		key_addrs->v4addrs.src = flow_keys->ipv4_src;
956 		key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
957 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
958 	} else if (flow_keys->addr_proto == ETH_P_IPV6 &&
959 		   dissector_uses_key(flow_dissector,
960 				      FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
961 		key_addrs = skb_flow_dissector_target(flow_dissector,
962 						      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
963 						      target_container);
964 		memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
965 		       sizeof(key_addrs->v6addrs.src));
966 		memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
967 		       sizeof(key_addrs->v6addrs.dst));
968 		key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
969 	}
970 
971 	if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
972 		key_ports = skb_flow_dissector_target(flow_dissector,
973 						      FLOW_DISSECTOR_KEY_PORTS,
974 						      target_container);
975 	else if (dissector_uses_key(flow_dissector,
976 				    FLOW_DISSECTOR_KEY_PORTS_RANGE))
977 		key_ports = skb_flow_dissector_target(flow_dissector,
978 						      FLOW_DISSECTOR_KEY_PORTS_RANGE,
979 						      target_container);
980 
981 	if (key_ports) {
982 		key_ports->src = flow_keys->sport;
983 		key_ports->dst = flow_keys->dport;
984 	}
985 
986 	if (dissector_uses_key(flow_dissector,
987 			       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
988 		key_tags = skb_flow_dissector_target(flow_dissector,
989 						     FLOW_DISSECTOR_KEY_FLOW_LABEL,
990 						     target_container);
991 		key_tags->flow_label = ntohl(flow_keys->flow_label);
992 	}
993 }
994 
bpf_flow_dissect(struct bpf_prog * prog,struct bpf_flow_dissector * ctx,__be16 proto,int nhoff,int hlen,unsigned int flags)995 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
996 		     __be16 proto, int nhoff, int hlen, unsigned int flags)
997 {
998 	struct bpf_flow_keys *flow_keys = ctx->flow_keys;
999 	u32 result;
1000 
1001 	/* Pass parameters to the BPF program */
1002 	memset(flow_keys, 0, sizeof(*flow_keys));
1003 	flow_keys->n_proto = proto;
1004 	flow_keys->nhoff = nhoff;
1005 	flow_keys->thoff = flow_keys->nhoff;
1006 
1007 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
1008 		     (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
1009 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
1010 		     (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1011 	BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
1012 		     (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
1013 	flow_keys->flags = flags;
1014 
1015 	result = bpf_prog_run_pin_on_cpu(prog, ctx);
1016 
1017 	flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1018 	flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1019 				   flow_keys->nhoff, hlen);
1020 
1021 	return result;
1022 }
1023 
is_pppoe_ses_hdr_valid(const struct pppoe_hdr * hdr)1024 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1025 {
1026 	return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1027 }
1028 
1029 /**
1030  * __skb_flow_dissect - extract the flow_keys struct and return it
1031  * @net: associated network namespace, derived from @skb if NULL
1032  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1033  * @flow_dissector: list of keys to dissect
1034  * @target_container: target structure to put dissected values into
1035  * @data: raw buffer pointer to the packet, if NULL use skb->data
1036  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1037  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1038  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1039  * @flags: flags that control the dissection process, e.g.
1040  *         FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1041  *
1042  * The function will try to retrieve individual keys into target specified
1043  * by flow_dissector from either the skbuff or a raw buffer specified by the
1044  * rest parameters.
1045  *
1046  * Caller must take care of zeroing target container memory.
1047  */
__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)1048 bool __skb_flow_dissect(const struct net *net,
1049 			const struct sk_buff *skb,
1050 			struct flow_dissector *flow_dissector,
1051 			void *target_container, const void *data,
1052 			__be16 proto, int nhoff, int hlen, unsigned int flags)
1053 {
1054 	struct flow_dissector_key_control *key_control;
1055 	struct flow_dissector_key_basic *key_basic;
1056 	struct flow_dissector_key_addrs *key_addrs;
1057 	struct flow_dissector_key_tags *key_tags;
1058 	struct flow_dissector_key_vlan *key_vlan;
1059 	enum flow_dissect_ret fdret;
1060 	enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1061 	bool mpls_el = false;
1062 	int mpls_lse = 0;
1063 	int num_hdrs = 0;
1064 	u8 ip_proto = 0;
1065 	bool ret;
1066 
1067 	if (!data) {
1068 		data = skb->data;
1069 		proto = skb_vlan_tag_present(skb) ?
1070 			 skb->vlan_proto : skb->protocol;
1071 		nhoff = skb_network_offset(skb);
1072 		hlen = skb_headlen(skb);
1073 #if IS_ENABLED(CONFIG_NET_DSA)
1074 		if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1075 			     proto == htons(ETH_P_XDSA))) {
1076 			struct metadata_dst *md_dst = skb_metadata_dst(skb);
1077 			const struct dsa_device_ops *ops;
1078 			int offset = 0;
1079 
1080 			ops = skb->dev->dsa_ptr->tag_ops;
1081 			/* Only DSA header taggers break flow dissection */
1082 			if (ops->needed_headroom &&
1083 			    (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1084 				if (ops->flow_dissect)
1085 					ops->flow_dissect(skb, &proto, &offset);
1086 				else
1087 					dsa_tag_generic_flow_dissect(skb,
1088 								     &proto,
1089 								     &offset);
1090 				hlen -= offset;
1091 				nhoff += offset;
1092 			}
1093 		}
1094 #endif
1095 	}
1096 
1097 	/* It is ensured by skb_flow_dissector_init() that control key will
1098 	 * be always present.
1099 	 */
1100 	key_control = skb_flow_dissector_target(flow_dissector,
1101 						FLOW_DISSECTOR_KEY_CONTROL,
1102 						target_container);
1103 
1104 	/* It is ensured by skb_flow_dissector_init() that basic key will
1105 	 * be always present.
1106 	 */
1107 	key_basic = skb_flow_dissector_target(flow_dissector,
1108 					      FLOW_DISSECTOR_KEY_BASIC,
1109 					      target_container);
1110 
1111 	if (skb) {
1112 		if (!net) {
1113 			if (skb->dev)
1114 				net = dev_net(skb->dev);
1115 			else if (skb->sk)
1116 				net = sock_net(skb->sk);
1117 		}
1118 	}
1119 
1120 	DEBUG_NET_WARN_ON_ONCE(!net);
1121 	if (net) {
1122 		enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1123 		struct bpf_prog_array *run_array;
1124 
1125 		rcu_read_lock();
1126 		run_array = rcu_dereference(init_net.bpf.run_array[type]);
1127 		if (!run_array)
1128 			run_array = rcu_dereference(net->bpf.run_array[type]);
1129 
1130 		if (run_array) {
1131 			struct bpf_flow_keys flow_keys;
1132 			struct bpf_flow_dissector ctx = {
1133 				.flow_keys = &flow_keys,
1134 				.data = data,
1135 				.data_end = data + hlen,
1136 			};
1137 			__be16 n_proto = proto;
1138 			struct bpf_prog *prog;
1139 			u32 result;
1140 
1141 			if (skb) {
1142 				ctx.skb = skb;
1143 				/* we can't use 'proto' in the skb case
1144 				 * because it might be set to skb->vlan_proto
1145 				 * which has been pulled from the data
1146 				 */
1147 				n_proto = skb->protocol;
1148 			}
1149 
1150 			prog = READ_ONCE(run_array->items[0].prog);
1151 			result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1152 						  hlen, flags);
1153 			if (result == BPF_FLOW_DISSECTOR_CONTINUE)
1154 				goto dissect_continue;
1155 			__skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1156 						 target_container);
1157 			rcu_read_unlock();
1158 			return result == BPF_OK;
1159 		}
1160 dissect_continue:
1161 		rcu_read_unlock();
1162 	}
1163 
1164 	if (dissector_uses_key(flow_dissector,
1165 			       FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1166 		struct ethhdr *eth = eth_hdr(skb);
1167 		struct flow_dissector_key_eth_addrs *key_eth_addrs;
1168 
1169 		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1170 							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
1171 							  target_container);
1172 		memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1173 	}
1174 
1175 	if (dissector_uses_key(flow_dissector,
1176 			       FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1177 		struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1178 
1179 		key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1180 							     FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1181 							     target_container);
1182 		key_num_of_vlans->num_of_vlans = 0;
1183 	}
1184 
1185 proto_again:
1186 	fdret = FLOW_DISSECT_RET_CONTINUE;
1187 
1188 	switch (proto) {
1189 	case htons(ETH_P_IP): {
1190 		const struct iphdr *iph;
1191 		struct iphdr _iph;
1192 
1193 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1194 		if (!iph || iph->ihl < 5) {
1195 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1196 			break;
1197 		}
1198 
1199 		nhoff += iph->ihl * 4;
1200 
1201 		ip_proto = iph->protocol;
1202 
1203 		if (dissector_uses_key(flow_dissector,
1204 				       FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1205 			key_addrs = skb_flow_dissector_target(flow_dissector,
1206 							      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1207 							      target_container);
1208 
1209 			memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1210 			       sizeof(key_addrs->v4addrs.src));
1211 			memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1212 			       sizeof(key_addrs->v4addrs.dst));
1213 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1214 		}
1215 
1216 		__skb_flow_dissect_ipv4(skb, flow_dissector,
1217 					target_container, data, iph);
1218 
1219 		if (ip_is_fragment(iph)) {
1220 			key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1221 
1222 			if (iph->frag_off & htons(IP_OFFSET)) {
1223 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1224 				break;
1225 			} else {
1226 				key_control->flags |= FLOW_DIS_FIRST_FRAG;
1227 				if (!(flags &
1228 				      FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1229 					fdret = FLOW_DISSECT_RET_OUT_GOOD;
1230 					break;
1231 				}
1232 			}
1233 		}
1234 
1235 		break;
1236 	}
1237 	case htons(ETH_P_IPV6): {
1238 		const struct ipv6hdr *iph;
1239 		struct ipv6hdr _iph;
1240 
1241 		iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1242 		if (!iph) {
1243 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1244 			break;
1245 		}
1246 
1247 		ip_proto = iph->nexthdr;
1248 		nhoff += sizeof(struct ipv6hdr);
1249 
1250 		if (dissector_uses_key(flow_dissector,
1251 				       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1252 			key_addrs = skb_flow_dissector_target(flow_dissector,
1253 							      FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1254 							      target_container);
1255 
1256 			memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1257 			       sizeof(key_addrs->v6addrs.src));
1258 			memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1259 			       sizeof(key_addrs->v6addrs.dst));
1260 			key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1261 		}
1262 
1263 		if ((dissector_uses_key(flow_dissector,
1264 					FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1265 		     (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1266 		    ip6_flowlabel(iph)) {
1267 			__be32 flow_label = ip6_flowlabel(iph);
1268 
1269 			if (dissector_uses_key(flow_dissector,
1270 					       FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1271 				key_tags = skb_flow_dissector_target(flow_dissector,
1272 								     FLOW_DISSECTOR_KEY_FLOW_LABEL,
1273 								     target_container);
1274 				key_tags->flow_label = ntohl(flow_label);
1275 			}
1276 			if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1277 				fdret = FLOW_DISSECT_RET_OUT_GOOD;
1278 				break;
1279 			}
1280 		}
1281 
1282 		__skb_flow_dissect_ipv6(skb, flow_dissector,
1283 					target_container, data, iph);
1284 
1285 		break;
1286 	}
1287 	case htons(ETH_P_8021AD):
1288 	case htons(ETH_P_8021Q): {
1289 		const struct vlan_hdr *vlan = NULL;
1290 		struct vlan_hdr _vlan;
1291 		__be16 saved_vlan_tpid = proto;
1292 
1293 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1294 		    skb && skb_vlan_tag_present(skb)) {
1295 			proto = skb->protocol;
1296 		} else {
1297 			vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1298 						    data, hlen, &_vlan);
1299 			if (!vlan) {
1300 				fdret = FLOW_DISSECT_RET_OUT_BAD;
1301 				break;
1302 			}
1303 
1304 			proto = vlan->h_vlan_encapsulated_proto;
1305 			nhoff += sizeof(*vlan);
1306 		}
1307 
1308 		if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1309 		    !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1310 			struct flow_dissector_key_num_of_vlans *key_nvs;
1311 
1312 			key_nvs = skb_flow_dissector_target(flow_dissector,
1313 							    FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1314 							    target_container);
1315 			key_nvs->num_of_vlans++;
1316 		}
1317 
1318 		if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1319 			dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1320 		} else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1321 			dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1322 		} else {
1323 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1324 			break;
1325 		}
1326 
1327 		if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1328 			key_vlan = skb_flow_dissector_target(flow_dissector,
1329 							     dissector_vlan,
1330 							     target_container);
1331 
1332 			if (!vlan) {
1333 				key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1334 				key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1335 			} else {
1336 				key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1337 					VLAN_VID_MASK;
1338 				key_vlan->vlan_priority =
1339 					(ntohs(vlan->h_vlan_TCI) &
1340 					 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1341 			}
1342 			key_vlan->vlan_tpid = saved_vlan_tpid;
1343 			key_vlan->vlan_eth_type = proto;
1344 		}
1345 
1346 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1347 		break;
1348 	}
1349 	case htons(ETH_P_PPP_SES): {
1350 		struct {
1351 			struct pppoe_hdr hdr;
1352 			__be16 proto;
1353 		} *hdr, _hdr;
1354 		u16 ppp_proto;
1355 
1356 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1357 		if (!hdr) {
1358 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1359 			break;
1360 		}
1361 
1362 		if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1363 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1364 			break;
1365 		}
1366 
1367 		/* least significant bit of the most significant octet
1368 		 * indicates if protocol field was compressed
1369 		 */
1370 		ppp_proto = ntohs(hdr->proto);
1371 		if (ppp_proto & 0x0100) {
1372 			ppp_proto = ppp_proto >> 8;
1373 			nhoff += PPPOE_SES_HLEN - 1;
1374 		} else {
1375 			nhoff += PPPOE_SES_HLEN;
1376 		}
1377 
1378 		if (ppp_proto == PPP_IP) {
1379 			proto = htons(ETH_P_IP);
1380 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1381 		} else if (ppp_proto == PPP_IPV6) {
1382 			proto = htons(ETH_P_IPV6);
1383 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1384 		} else if (ppp_proto == PPP_MPLS_UC) {
1385 			proto = htons(ETH_P_MPLS_UC);
1386 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1387 		} else if (ppp_proto == PPP_MPLS_MC) {
1388 			proto = htons(ETH_P_MPLS_MC);
1389 			fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1390 		} else if (ppp_proto_is_valid(ppp_proto)) {
1391 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1392 		} else {
1393 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1394 			break;
1395 		}
1396 
1397 		if (dissector_uses_key(flow_dissector,
1398 				       FLOW_DISSECTOR_KEY_PPPOE)) {
1399 			struct flow_dissector_key_pppoe *key_pppoe;
1400 
1401 			key_pppoe = skb_flow_dissector_target(flow_dissector,
1402 							      FLOW_DISSECTOR_KEY_PPPOE,
1403 							      target_container);
1404 			key_pppoe->session_id = hdr->hdr.sid;
1405 			key_pppoe->ppp_proto = htons(ppp_proto);
1406 			key_pppoe->type = htons(ETH_P_PPP_SES);
1407 		}
1408 		break;
1409 	}
1410 	case htons(ETH_P_TIPC): {
1411 		struct tipc_basic_hdr *hdr, _hdr;
1412 
1413 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1414 					   data, hlen, &_hdr);
1415 		if (!hdr) {
1416 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1417 			break;
1418 		}
1419 
1420 		if (dissector_uses_key(flow_dissector,
1421 				       FLOW_DISSECTOR_KEY_TIPC)) {
1422 			key_addrs = skb_flow_dissector_target(flow_dissector,
1423 							      FLOW_DISSECTOR_KEY_TIPC,
1424 							      target_container);
1425 			key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1426 			key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1427 		}
1428 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1429 		break;
1430 	}
1431 
1432 	case htons(ETH_P_MPLS_UC):
1433 	case htons(ETH_P_MPLS_MC):
1434 		fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1435 						target_container, data,
1436 						nhoff, hlen, mpls_lse,
1437 						&mpls_el);
1438 		nhoff += sizeof(struct mpls_label);
1439 		mpls_lse++;
1440 		break;
1441 	case htons(ETH_P_FCOE):
1442 		if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1443 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1444 			break;
1445 		}
1446 
1447 		nhoff += FCOE_HEADER_LEN;
1448 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1449 		break;
1450 
1451 	case htons(ETH_P_ARP):
1452 	case htons(ETH_P_RARP):
1453 		fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1454 					       target_container, data,
1455 					       nhoff, hlen);
1456 		break;
1457 
1458 	case htons(ETH_P_BATMAN):
1459 		fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1460 						  &proto, &nhoff, hlen, flags);
1461 		break;
1462 
1463 	case htons(ETH_P_1588): {
1464 		struct ptp_header *hdr, _hdr;
1465 
1466 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1467 					   hlen, &_hdr);
1468 		if (!hdr) {
1469 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1470 			break;
1471 		}
1472 
1473 		nhoff += sizeof(struct ptp_header);
1474 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1475 		break;
1476 	}
1477 
1478 	case htons(ETH_P_PRP):
1479 	case htons(ETH_P_HSR): {
1480 		struct hsr_tag *hdr, _hdr;
1481 
1482 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1483 					   &_hdr);
1484 		if (!hdr) {
1485 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1486 			break;
1487 		}
1488 
1489 		proto = hdr->encap_proto;
1490 		nhoff += HSR_HLEN;
1491 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1492 		break;
1493 	}
1494 
1495 	case htons(ETH_P_CFM):
1496 		fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1497 					       target_container, data,
1498 					       nhoff, hlen);
1499 		break;
1500 
1501 	default:
1502 		fdret = FLOW_DISSECT_RET_OUT_BAD;
1503 		break;
1504 	}
1505 
1506 	/* Process result of proto processing */
1507 	switch (fdret) {
1508 	case FLOW_DISSECT_RET_OUT_GOOD:
1509 		goto out_good;
1510 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1511 		if (skb_flow_dissect_allowed(&num_hdrs))
1512 			goto proto_again;
1513 		goto out_good;
1514 	case FLOW_DISSECT_RET_CONTINUE:
1515 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1516 		break;
1517 	case FLOW_DISSECT_RET_OUT_BAD:
1518 	default:
1519 		goto out_bad;
1520 	}
1521 
1522 ip_proto_again:
1523 	fdret = FLOW_DISSECT_RET_CONTINUE;
1524 
1525 	switch (ip_proto) {
1526 	case IPPROTO_GRE:
1527 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1528 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1529 			break;
1530 		}
1531 
1532 		fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1533 					       target_container, data,
1534 					       &proto, &nhoff, &hlen, flags);
1535 		break;
1536 
1537 	case NEXTHDR_HOP:
1538 	case NEXTHDR_ROUTING:
1539 	case NEXTHDR_DEST: {
1540 		u8 _opthdr[2], *opthdr;
1541 
1542 		if (proto != htons(ETH_P_IPV6))
1543 			break;
1544 
1545 		opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1546 					      data, hlen, &_opthdr);
1547 		if (!opthdr) {
1548 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1549 			break;
1550 		}
1551 
1552 		ip_proto = opthdr[0];
1553 		nhoff += (opthdr[1] + 1) << 3;
1554 
1555 		fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1556 		break;
1557 	}
1558 	case NEXTHDR_FRAGMENT: {
1559 		struct frag_hdr _fh, *fh;
1560 
1561 		if (proto != htons(ETH_P_IPV6))
1562 			break;
1563 
1564 		fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1565 					  data, hlen, &_fh);
1566 
1567 		if (!fh) {
1568 			fdret = FLOW_DISSECT_RET_OUT_BAD;
1569 			break;
1570 		}
1571 
1572 		key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1573 
1574 		nhoff += sizeof(_fh);
1575 		ip_proto = fh->nexthdr;
1576 
1577 		if (!(fh->frag_off & htons(IP6_OFFSET))) {
1578 			key_control->flags |= FLOW_DIS_FIRST_FRAG;
1579 			if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1580 				fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1581 				break;
1582 			}
1583 		}
1584 
1585 		fdret = FLOW_DISSECT_RET_OUT_GOOD;
1586 		break;
1587 	}
1588 	case IPPROTO_IPIP:
1589 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1590 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1591 			break;
1592 		}
1593 
1594 		proto = htons(ETH_P_IP);
1595 
1596 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1597 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1598 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1599 			break;
1600 		}
1601 
1602 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1603 		break;
1604 
1605 	case IPPROTO_IPV6:
1606 		if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1607 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1608 			break;
1609 		}
1610 
1611 		proto = htons(ETH_P_IPV6);
1612 
1613 		key_control->flags |= FLOW_DIS_ENCAPSULATION;
1614 		if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1615 			fdret = FLOW_DISSECT_RET_OUT_GOOD;
1616 			break;
1617 		}
1618 
1619 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1620 		break;
1621 
1622 
1623 	case IPPROTO_MPLS:
1624 		proto = htons(ETH_P_MPLS_UC);
1625 		fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1626 		break;
1627 
1628 	case IPPROTO_TCP:
1629 		__skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1630 				       data, nhoff, hlen);
1631 		break;
1632 
1633 	case IPPROTO_ICMP:
1634 	case IPPROTO_ICMPV6:
1635 		__skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1636 					data, nhoff, hlen);
1637 		break;
1638 	case IPPROTO_L2TP:
1639 		__skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1640 					  data, nhoff, hlen);
1641 		break;
1642 	case IPPROTO_ESP:
1643 		__skb_flow_dissect_esp(skb, flow_dissector, target_container,
1644 				       data, nhoff, hlen);
1645 		break;
1646 	case IPPROTO_AH:
1647 		__skb_flow_dissect_ah(skb, flow_dissector, target_container,
1648 				      data, nhoff, hlen);
1649 		break;
1650 	default:
1651 		break;
1652 	}
1653 
1654 	if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1655 		__skb_flow_dissect_ports(skb, flow_dissector, target_container,
1656 					 data, nhoff, ip_proto, hlen);
1657 
1658 	/* Process result of IP proto processing */
1659 	switch (fdret) {
1660 	case FLOW_DISSECT_RET_PROTO_AGAIN:
1661 		if (skb_flow_dissect_allowed(&num_hdrs))
1662 			goto proto_again;
1663 		break;
1664 	case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1665 		if (skb_flow_dissect_allowed(&num_hdrs))
1666 			goto ip_proto_again;
1667 		break;
1668 	case FLOW_DISSECT_RET_OUT_GOOD:
1669 	case FLOW_DISSECT_RET_CONTINUE:
1670 		break;
1671 	case FLOW_DISSECT_RET_OUT_BAD:
1672 	default:
1673 		goto out_bad;
1674 	}
1675 
1676 out_good:
1677 	ret = true;
1678 
1679 out:
1680 	key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1681 	key_basic->n_proto = proto;
1682 	key_basic->ip_proto = ip_proto;
1683 
1684 	return ret;
1685 
1686 out_bad:
1687 	ret = false;
1688 	goto out;
1689 }
1690 EXPORT_SYMBOL(__skb_flow_dissect);
1691 
1692 static siphash_aligned_key_t hashrnd;
__flow_hash_secret_init(void)1693 static __always_inline void __flow_hash_secret_init(void)
1694 {
1695 	net_get_random_once(&hashrnd, sizeof(hashrnd));
1696 }
1697 
flow_keys_hash_start(const struct flow_keys * flow)1698 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1699 {
1700 	BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1701 	return &flow->FLOW_KEYS_HASH_START_FIELD;
1702 }
1703 
flow_keys_hash_length(const struct flow_keys * flow)1704 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1705 {
1706 	size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1707 
1708 	BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1709 
1710 	switch (flow->control.addr_type) {
1711 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1712 		diff -= sizeof(flow->addrs.v4addrs);
1713 		break;
1714 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1715 		diff -= sizeof(flow->addrs.v6addrs);
1716 		break;
1717 	case FLOW_DISSECTOR_KEY_TIPC:
1718 		diff -= sizeof(flow->addrs.tipckey);
1719 		break;
1720 	}
1721 	return sizeof(*flow) - diff;
1722 }
1723 
flow_get_u32_src(const struct flow_keys * flow)1724 __be32 flow_get_u32_src(const struct flow_keys *flow)
1725 {
1726 	switch (flow->control.addr_type) {
1727 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1728 		return flow->addrs.v4addrs.src;
1729 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1730 		return (__force __be32)ipv6_addr_hash(
1731 			&flow->addrs.v6addrs.src);
1732 	case FLOW_DISSECTOR_KEY_TIPC:
1733 		return flow->addrs.tipckey.key;
1734 	default:
1735 		return 0;
1736 	}
1737 }
1738 EXPORT_SYMBOL(flow_get_u32_src);
1739 
flow_get_u32_dst(const struct flow_keys * flow)1740 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1741 {
1742 	switch (flow->control.addr_type) {
1743 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1744 		return flow->addrs.v4addrs.dst;
1745 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1746 		return (__force __be32)ipv6_addr_hash(
1747 			&flow->addrs.v6addrs.dst);
1748 	default:
1749 		return 0;
1750 	}
1751 }
1752 EXPORT_SYMBOL(flow_get_u32_dst);
1753 
1754 /* Sort the source and destination IP and the ports,
1755  * to have consistent hash within the two directions
1756  */
__flow_hash_consistentify(struct flow_keys * keys)1757 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1758 {
1759 	int addr_diff, i;
1760 
1761 	switch (keys->control.addr_type) {
1762 	case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1763 		if ((__force u32)keys->addrs.v4addrs.dst <
1764 		    (__force u32)keys->addrs.v4addrs.src)
1765 			swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1766 
1767 		if ((__force u16)keys->ports.dst <
1768 		    (__force u16)keys->ports.src) {
1769 			swap(keys->ports.src, keys->ports.dst);
1770 		}
1771 		break;
1772 	case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1773 		addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1774 				   &keys->addrs.v6addrs.src,
1775 				   sizeof(keys->addrs.v6addrs.dst));
1776 		if (addr_diff < 0) {
1777 			for (i = 0; i < 4; i++)
1778 				swap(keys->addrs.v6addrs.src.s6_addr32[i],
1779 				     keys->addrs.v6addrs.dst.s6_addr32[i]);
1780 		}
1781 		if ((__force u16)keys->ports.dst <
1782 		    (__force u16)keys->ports.src) {
1783 			swap(keys->ports.src, keys->ports.dst);
1784 		}
1785 		break;
1786 	}
1787 }
1788 
__flow_hash_from_keys(struct flow_keys * keys,const siphash_key_t * keyval)1789 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1790 					const siphash_key_t *keyval)
1791 {
1792 	u32 hash;
1793 
1794 	__flow_hash_consistentify(keys);
1795 
1796 	hash = siphash(flow_keys_hash_start(keys),
1797 		       flow_keys_hash_length(keys), keyval);
1798 	if (!hash)
1799 		hash = 1;
1800 
1801 	return hash;
1802 }
1803 
flow_hash_from_keys(struct flow_keys * keys)1804 u32 flow_hash_from_keys(struct flow_keys *keys)
1805 {
1806 	__flow_hash_secret_init();
1807 	return __flow_hash_from_keys(keys, &hashrnd);
1808 }
1809 EXPORT_SYMBOL(flow_hash_from_keys);
1810 
flow_hash_from_keys_seed(struct flow_keys * keys,const siphash_key_t * keyval)1811 u32 flow_hash_from_keys_seed(struct flow_keys *keys,
1812 			     const siphash_key_t *keyval)
1813 {
1814 	return __flow_hash_from_keys(keys, keyval);
1815 }
1816 EXPORT_SYMBOL(flow_hash_from_keys_seed);
1817 
___skb_get_hash(const struct sk_buff * skb,struct flow_keys * keys,const siphash_key_t * keyval)1818 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1819 				  struct flow_keys *keys,
1820 				  const siphash_key_t *keyval)
1821 {
1822 	skb_flow_dissect_flow_keys(skb, keys,
1823 				   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1824 
1825 	return __flow_hash_from_keys(keys, keyval);
1826 }
1827 
1828 struct _flow_keys_digest_data {
1829 	__be16	n_proto;
1830 	u8	ip_proto;
1831 	u8	padding;
1832 	__be32	ports;
1833 	__be32	src;
1834 	__be32	dst;
1835 };
1836 
make_flow_keys_digest(struct flow_keys_digest * digest,const struct flow_keys * flow)1837 void make_flow_keys_digest(struct flow_keys_digest *digest,
1838 			   const struct flow_keys *flow)
1839 {
1840 	struct _flow_keys_digest_data *data =
1841 	    (struct _flow_keys_digest_data *)digest;
1842 
1843 	BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1844 
1845 	memset(digest, 0, sizeof(*digest));
1846 
1847 	data->n_proto = flow->basic.n_proto;
1848 	data->ip_proto = flow->basic.ip_proto;
1849 	data->ports = flow->ports.ports;
1850 	data->src = flow->addrs.v4addrs.src;
1851 	data->dst = flow->addrs.v4addrs.dst;
1852 }
1853 EXPORT_SYMBOL(make_flow_keys_digest);
1854 
1855 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1856 
__skb_get_hash_symmetric_net(const struct net * net,const struct sk_buff * skb)1857 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb)
1858 {
1859 	struct flow_keys keys;
1860 
1861 	__flow_hash_secret_init();
1862 
1863 	memset(&keys, 0, sizeof(keys));
1864 	__skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric,
1865 			   &keys, NULL, 0, 0, 0, 0);
1866 
1867 	return __flow_hash_from_keys(&keys, &hashrnd);
1868 }
1869 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net);
1870 
1871 /**
1872  * __skb_get_hash_net: calculate a flow hash
1873  * @net: associated network namespace, derived from @skb if NULL
1874  * @skb: sk_buff to calculate flow hash from
1875  *
1876  * This function calculates a flow hash based on src/dst addresses
1877  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1878  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1879  * if hash is a canonical 4-tuple hash over transport ports.
1880  */
__skb_get_hash_net(const struct net * net,struct sk_buff * skb)1881 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1882 {
1883 	struct flow_keys keys;
1884 	u32 hash;
1885 
1886 	memset(&keys, 0, sizeof(keys));
1887 
1888 	__skb_flow_dissect(net, skb, &flow_keys_dissector,
1889 			   &keys, NULL, 0, 0, 0,
1890 			   FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1891 
1892 	__flow_hash_secret_init();
1893 
1894 	hash = __flow_hash_from_keys(&keys, &hashrnd);
1895 
1896 	__skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1897 }
1898 EXPORT_SYMBOL(__skb_get_hash_net);
1899 
skb_get_hash_perturb(const struct sk_buff * skb,const siphash_key_t * perturb)1900 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1901 			   const siphash_key_t *perturb)
1902 {
1903 	struct flow_keys keys;
1904 
1905 	return ___skb_get_hash(skb, &keys, perturb);
1906 }
1907 EXPORT_SYMBOL(skb_get_hash_perturb);
1908 
__skb_get_poff(const struct sk_buff * skb,const void * data,const struct flow_keys_basic * keys,int hlen)1909 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1910 		   const struct flow_keys_basic *keys, int hlen)
1911 {
1912 	u32 poff = keys->control.thoff;
1913 
1914 	/* skip L4 headers for fragments after the first */
1915 	if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1916 	    !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1917 		return poff;
1918 
1919 	switch (keys->basic.ip_proto) {
1920 	case IPPROTO_TCP: {
1921 		/* access doff as u8 to avoid unaligned access */
1922 		const u8 *doff;
1923 		u8 _doff;
1924 
1925 		doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1926 					    data, hlen, &_doff);
1927 		if (!doff)
1928 			return poff;
1929 
1930 		poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1931 		break;
1932 	}
1933 	case IPPROTO_UDP:
1934 	case IPPROTO_UDPLITE:
1935 		poff += sizeof(struct udphdr);
1936 		break;
1937 	/* For the rest, we do not really care about header
1938 	 * extensions at this point for now.
1939 	 */
1940 	case IPPROTO_ICMP:
1941 		poff += sizeof(struct icmphdr);
1942 		break;
1943 	case IPPROTO_ICMPV6:
1944 		poff += sizeof(struct icmp6hdr);
1945 		break;
1946 	case IPPROTO_IGMP:
1947 		poff += sizeof(struct igmphdr);
1948 		break;
1949 	case IPPROTO_DCCP:
1950 		poff += sizeof(struct dccp_hdr);
1951 		break;
1952 	case IPPROTO_SCTP:
1953 		poff += sizeof(struct sctphdr);
1954 		break;
1955 	}
1956 
1957 	return poff;
1958 }
1959 
1960 /**
1961  * skb_get_poff - get the offset to the payload
1962  * @skb: sk_buff to get the payload offset from
1963  *
1964  * The function will get the offset to the payload as far as it could
1965  * be dissected.  The main user is currently BPF, so that we can dynamically
1966  * truncate packets without needing to push actual payload to the user
1967  * space and can analyze headers only, instead.
1968  */
skb_get_poff(const struct sk_buff * skb)1969 u32 skb_get_poff(const struct sk_buff *skb)
1970 {
1971 	struct flow_keys_basic keys;
1972 
1973 	if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1974 					      NULL, 0, 0, 0, 0))
1975 		return 0;
1976 
1977 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1978 }
1979 
__get_hash_from_flowi6(const struct flowi6 * fl6,struct flow_keys * keys)1980 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1981 {
1982 	memset(keys, 0, sizeof(*keys));
1983 
1984 	memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1985 	    sizeof(keys->addrs.v6addrs.src));
1986 	memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1987 	    sizeof(keys->addrs.v6addrs.dst));
1988 	keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1989 	keys->ports.src = fl6->fl6_sport;
1990 	keys->ports.dst = fl6->fl6_dport;
1991 	keys->keyid.keyid = fl6->fl6_gre_key;
1992 	keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1993 	keys->basic.ip_proto = fl6->flowi6_proto;
1994 
1995 	return flow_hash_from_keys(keys);
1996 }
1997 EXPORT_SYMBOL(__get_hash_from_flowi6);
1998 
1999 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
2000 	{
2001 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2002 		.offset = offsetof(struct flow_keys, control),
2003 	},
2004 	{
2005 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2006 		.offset = offsetof(struct flow_keys, basic),
2007 	},
2008 	{
2009 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2010 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
2011 	},
2012 	{
2013 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2014 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
2015 	},
2016 	{
2017 		.key_id = FLOW_DISSECTOR_KEY_TIPC,
2018 		.offset = offsetof(struct flow_keys, addrs.tipckey),
2019 	},
2020 	{
2021 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
2022 		.offset = offsetof(struct flow_keys, ports),
2023 	},
2024 	{
2025 		.key_id = FLOW_DISSECTOR_KEY_VLAN,
2026 		.offset = offsetof(struct flow_keys, vlan),
2027 	},
2028 	{
2029 		.key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2030 		.offset = offsetof(struct flow_keys, tags),
2031 	},
2032 	{
2033 		.key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2034 		.offset = offsetof(struct flow_keys, keyid),
2035 	},
2036 };
2037 
2038 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2039 	{
2040 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2041 		.offset = offsetof(struct flow_keys, control),
2042 	},
2043 	{
2044 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2045 		.offset = offsetof(struct flow_keys, basic),
2046 	},
2047 	{
2048 		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2049 		.offset = offsetof(struct flow_keys, addrs.v4addrs),
2050 	},
2051 	{
2052 		.key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2053 		.offset = offsetof(struct flow_keys, addrs.v6addrs),
2054 	},
2055 	{
2056 		.key_id = FLOW_DISSECTOR_KEY_PORTS,
2057 		.offset = offsetof(struct flow_keys, ports),
2058 	},
2059 };
2060 
2061 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2062 	{
2063 		.key_id = FLOW_DISSECTOR_KEY_CONTROL,
2064 		.offset = offsetof(struct flow_keys, control),
2065 	},
2066 	{
2067 		.key_id = FLOW_DISSECTOR_KEY_BASIC,
2068 		.offset = offsetof(struct flow_keys, basic),
2069 	},
2070 };
2071 
2072 struct flow_dissector flow_keys_dissector __read_mostly;
2073 EXPORT_SYMBOL(flow_keys_dissector);
2074 
2075 struct flow_dissector flow_keys_basic_dissector __read_mostly;
2076 EXPORT_SYMBOL(flow_keys_basic_dissector);
2077 
init_default_flow_dissectors(void)2078 static int __init init_default_flow_dissectors(void)
2079 {
2080 	skb_flow_dissector_init(&flow_keys_dissector,
2081 				flow_keys_dissector_keys,
2082 				ARRAY_SIZE(flow_keys_dissector_keys));
2083 	skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2084 				flow_keys_dissector_symmetric_keys,
2085 				ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2086 	skb_flow_dissector_init(&flow_keys_basic_dissector,
2087 				flow_keys_basic_dissector_keys,
2088 				ARRAY_SIZE(flow_keys_basic_dissector_keys));
2089 	return 0;
2090 }
2091 core_initcall(init_default_flow_dissectors);
2092