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