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 struct flow_dissector_key_ports_range *key_ports_range = NULL;
833 struct flow_dissector_key_ports *key_ports = NULL;
834 __be32 ports;
835
836 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
837 key_ports = skb_flow_dissector_target(flow_dissector,
838 FLOW_DISSECTOR_KEY_PORTS,
839 target_container);
840
841 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS_RANGE))
842 key_ports_range = skb_flow_dissector_target(flow_dissector,
843 FLOW_DISSECTOR_KEY_PORTS_RANGE,
844 target_container);
845
846 if (!key_ports && !key_ports_range)
847 return;
848
849 ports = __skb_flow_get_ports(skb, nhoff, ip_proto, data, hlen);
850
851 if (key_ports)
852 key_ports->ports = ports;
853
854 if (key_ports_range)
855 key_ports_range->tp.ports = ports;
856 }
857
858 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)859 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
860 struct flow_dissector *flow_dissector,
861 void *target_container, const void *data,
862 const struct iphdr *iph)
863 {
864 struct flow_dissector_key_ip *key_ip;
865
866 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
867 return;
868
869 key_ip = skb_flow_dissector_target(flow_dissector,
870 FLOW_DISSECTOR_KEY_IP,
871 target_container);
872 key_ip->tos = iph->tos;
873 key_ip->ttl = iph->ttl;
874 }
875
876 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)877 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
878 struct flow_dissector *flow_dissector,
879 void *target_container, const void *data,
880 const struct ipv6hdr *iph)
881 {
882 struct flow_dissector_key_ip *key_ip;
883
884 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
885 return;
886
887 key_ip = skb_flow_dissector_target(flow_dissector,
888 FLOW_DISSECTOR_KEY_IP,
889 target_container);
890 key_ip->tos = ipv6_get_dsfield(iph);
891 key_ip->ttl = iph->hop_limit;
892 }
893
894 /* Maximum number of protocol headers that can be parsed in
895 * __skb_flow_dissect
896 */
897 #define MAX_FLOW_DISSECT_HDRS 15
898
skb_flow_dissect_allowed(int * num_hdrs)899 static bool skb_flow_dissect_allowed(int *num_hdrs)
900 {
901 ++*num_hdrs;
902
903 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
904 }
905
__skb_flow_bpf_to_target(const struct bpf_flow_keys * flow_keys,struct flow_dissector * flow_dissector,void * target_container)906 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
907 struct flow_dissector *flow_dissector,
908 void *target_container)
909 {
910 struct flow_dissector_key_ports_range *key_ports_range = NULL;
911 struct flow_dissector_key_ports *key_ports = NULL;
912 struct flow_dissector_key_control *key_control;
913 struct flow_dissector_key_basic *key_basic;
914 struct flow_dissector_key_addrs *key_addrs;
915 struct flow_dissector_key_tags *key_tags;
916
917 key_control = skb_flow_dissector_target(flow_dissector,
918 FLOW_DISSECTOR_KEY_CONTROL,
919 target_container);
920 key_control->thoff = flow_keys->thoff;
921 if (flow_keys->is_frag)
922 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
923 if (flow_keys->is_first_frag)
924 key_control->flags |= FLOW_DIS_FIRST_FRAG;
925 if (flow_keys->is_encap)
926 key_control->flags |= FLOW_DIS_ENCAPSULATION;
927
928 key_basic = skb_flow_dissector_target(flow_dissector,
929 FLOW_DISSECTOR_KEY_BASIC,
930 target_container);
931 key_basic->n_proto = flow_keys->n_proto;
932 key_basic->ip_proto = flow_keys->ip_proto;
933
934 if (flow_keys->addr_proto == ETH_P_IP &&
935 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
936 key_addrs = skb_flow_dissector_target(flow_dissector,
937 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
938 target_container);
939 key_addrs->v4addrs.src = flow_keys->ipv4_src;
940 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
941 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
942 } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
943 dissector_uses_key(flow_dissector,
944 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
945 key_addrs = skb_flow_dissector_target(flow_dissector,
946 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
947 target_container);
948 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
949 sizeof(key_addrs->v6addrs.src));
950 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
951 sizeof(key_addrs->v6addrs.dst));
952 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
953 }
954
955 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) {
956 key_ports = skb_flow_dissector_target(flow_dissector,
957 FLOW_DISSECTOR_KEY_PORTS,
958 target_container);
959 key_ports->src = flow_keys->sport;
960 key_ports->dst = flow_keys->dport;
961 }
962 if (dissector_uses_key(flow_dissector,
963 FLOW_DISSECTOR_KEY_PORTS_RANGE)) {
964 key_ports_range = skb_flow_dissector_target(flow_dissector,
965 FLOW_DISSECTOR_KEY_PORTS_RANGE,
966 target_container);
967 key_ports_range->tp.src = flow_keys->sport;
968 key_ports_range->tp.dst = flow_keys->dport;
969 }
970
971 if (dissector_uses_key(flow_dissector,
972 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
973 key_tags = skb_flow_dissector_target(flow_dissector,
974 FLOW_DISSECTOR_KEY_FLOW_LABEL,
975 target_container);
976 key_tags->flow_label = ntohl(flow_keys->flow_label);
977 }
978 }
979
bpf_flow_dissect(struct bpf_prog * prog,struct bpf_flow_dissector * ctx,__be16 proto,int nhoff,int hlen,unsigned int flags)980 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
981 __be16 proto, int nhoff, int hlen, unsigned int flags)
982 {
983 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
984 u32 result;
985
986 /* Pass parameters to the BPF program */
987 memset(flow_keys, 0, sizeof(*flow_keys));
988 flow_keys->n_proto = proto;
989 flow_keys->nhoff = nhoff;
990 flow_keys->thoff = flow_keys->nhoff;
991
992 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
993 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
994 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
995 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
996 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
997 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
998 flow_keys->flags = flags;
999
1000 result = bpf_prog_run_pin_on_cpu(prog, ctx);
1001
1002 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1003 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1004 flow_keys->nhoff, hlen);
1005
1006 return result;
1007 }
1008
is_pppoe_ses_hdr_valid(const struct pppoe_hdr * hdr)1009 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1010 {
1011 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1012 }
1013
1014 /**
1015 * __skb_flow_dissect - extract the flow_keys struct and return it
1016 * @net: associated network namespace, derived from @skb if NULL
1017 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1018 * @flow_dissector: list of keys to dissect
1019 * @target_container: target structure to put dissected values into
1020 * @data: raw buffer pointer to the packet, if NULL use skb->data
1021 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1022 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1023 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1024 * @flags: flags that control the dissection process, e.g.
1025 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1026 *
1027 * The function will try to retrieve individual keys into target specified
1028 * by flow_dissector from either the skbuff or a raw buffer specified by the
1029 * rest parameters.
1030 *
1031 * Caller must take care of zeroing target container memory.
1032 */
__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)1033 bool __skb_flow_dissect(const struct net *net,
1034 const struct sk_buff *skb,
1035 struct flow_dissector *flow_dissector,
1036 void *target_container, const void *data,
1037 __be16 proto, int nhoff, int hlen, unsigned int flags)
1038 {
1039 struct flow_dissector_key_control *key_control;
1040 struct flow_dissector_key_basic *key_basic;
1041 struct flow_dissector_key_addrs *key_addrs;
1042 struct flow_dissector_key_tags *key_tags;
1043 struct flow_dissector_key_vlan *key_vlan;
1044 enum flow_dissect_ret fdret;
1045 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1046 bool mpls_el = false;
1047 int mpls_lse = 0;
1048 int num_hdrs = 0;
1049 u8 ip_proto = 0;
1050 bool ret;
1051
1052 if (!data) {
1053 data = skb->data;
1054 proto = skb_vlan_tag_present(skb) ?
1055 skb->vlan_proto : skb->protocol;
1056 nhoff = skb_network_offset(skb);
1057 hlen = skb_headlen(skb);
1058 #if IS_ENABLED(CONFIG_NET_DSA)
1059 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1060 proto == htons(ETH_P_XDSA))) {
1061 struct metadata_dst *md_dst = skb_metadata_dst(skb);
1062 const struct dsa_device_ops *ops;
1063 int offset = 0;
1064
1065 ops = skb->dev->dsa_ptr->tag_ops;
1066 /* Only DSA header taggers break flow dissection */
1067 if (ops->needed_headroom &&
1068 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1069 if (ops->flow_dissect)
1070 ops->flow_dissect(skb, &proto, &offset);
1071 else
1072 dsa_tag_generic_flow_dissect(skb,
1073 &proto,
1074 &offset);
1075 hlen -= offset;
1076 nhoff += offset;
1077 }
1078 }
1079 #endif
1080 }
1081
1082 /* It is ensured by skb_flow_dissector_init() that control key will
1083 * be always present.
1084 */
1085 key_control = skb_flow_dissector_target(flow_dissector,
1086 FLOW_DISSECTOR_KEY_CONTROL,
1087 target_container);
1088
1089 /* It is ensured by skb_flow_dissector_init() that basic key will
1090 * be always present.
1091 */
1092 key_basic = skb_flow_dissector_target(flow_dissector,
1093 FLOW_DISSECTOR_KEY_BASIC,
1094 target_container);
1095
1096 rcu_read_lock();
1097
1098 if (skb) {
1099 if (!net) {
1100 if (skb->dev)
1101 net = dev_net_rcu(skb->dev);
1102 else if (skb->sk)
1103 net = sock_net(skb->sk);
1104 }
1105 }
1106
1107 DEBUG_NET_WARN_ON_ONCE(!net);
1108 if (net) {
1109 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1110 struct bpf_prog_array *run_array;
1111
1112 run_array = rcu_dereference(init_net.bpf.run_array[type]);
1113 if (!run_array)
1114 run_array = rcu_dereference(net->bpf.run_array[type]);
1115
1116 if (run_array) {
1117 struct bpf_flow_keys flow_keys;
1118 struct bpf_flow_dissector ctx = {
1119 .flow_keys = &flow_keys,
1120 .data = data,
1121 .data_end = data + hlen,
1122 };
1123 __be16 n_proto = proto;
1124 struct bpf_prog *prog;
1125 u32 result;
1126
1127 if (skb) {
1128 ctx.skb = skb;
1129 /* we can't use 'proto' in the skb case
1130 * because it might be set to skb->vlan_proto
1131 * which has been pulled from the data
1132 */
1133 n_proto = skb->protocol;
1134 }
1135
1136 prog = READ_ONCE(run_array->items[0].prog);
1137 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1138 hlen, flags);
1139 if (result != BPF_FLOW_DISSECTOR_CONTINUE) {
1140 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1141 target_container);
1142 rcu_read_unlock();
1143 return result == BPF_OK;
1144 }
1145 }
1146 }
1147
1148 rcu_read_unlock();
1149
1150 if (dissector_uses_key(flow_dissector,
1151 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1152 struct ethhdr *eth = eth_hdr(skb);
1153 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1154
1155 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1156 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1157 target_container);
1158 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1159 }
1160
1161 if (dissector_uses_key(flow_dissector,
1162 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1163 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1164
1165 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1166 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1167 target_container);
1168 key_num_of_vlans->num_of_vlans = 0;
1169 }
1170
1171 proto_again:
1172 fdret = FLOW_DISSECT_RET_CONTINUE;
1173
1174 switch (proto) {
1175 case htons(ETH_P_IP): {
1176 const struct iphdr *iph;
1177 struct iphdr _iph;
1178
1179 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1180 if (!iph || iph->ihl < 5) {
1181 fdret = FLOW_DISSECT_RET_OUT_BAD;
1182 break;
1183 }
1184
1185 nhoff += iph->ihl * 4;
1186
1187 ip_proto = iph->protocol;
1188
1189 if (dissector_uses_key(flow_dissector,
1190 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1191 key_addrs = skb_flow_dissector_target(flow_dissector,
1192 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1193 target_container);
1194
1195 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1196 sizeof(key_addrs->v4addrs.src));
1197 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1198 sizeof(key_addrs->v4addrs.dst));
1199 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1200 }
1201
1202 __skb_flow_dissect_ipv4(skb, flow_dissector,
1203 target_container, data, iph);
1204
1205 if (ip_is_fragment(iph)) {
1206 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1207
1208 if (iph->frag_off & htons(IP_OFFSET)) {
1209 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1210 break;
1211 } else {
1212 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1213 if (!(flags &
1214 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1215 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1216 break;
1217 }
1218 }
1219 }
1220
1221 break;
1222 }
1223 case htons(ETH_P_IPV6): {
1224 const struct ipv6hdr *iph;
1225 struct ipv6hdr _iph;
1226
1227 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1228 if (!iph) {
1229 fdret = FLOW_DISSECT_RET_OUT_BAD;
1230 break;
1231 }
1232
1233 ip_proto = iph->nexthdr;
1234 nhoff += sizeof(struct ipv6hdr);
1235
1236 if (dissector_uses_key(flow_dissector,
1237 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1238 key_addrs = skb_flow_dissector_target(flow_dissector,
1239 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1240 target_container);
1241
1242 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1243 sizeof(key_addrs->v6addrs.src));
1244 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1245 sizeof(key_addrs->v6addrs.dst));
1246 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1247 }
1248
1249 if ((dissector_uses_key(flow_dissector,
1250 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1251 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1252 ip6_flowlabel(iph)) {
1253 __be32 flow_label = ip6_flowlabel(iph);
1254
1255 if (dissector_uses_key(flow_dissector,
1256 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1257 key_tags = skb_flow_dissector_target(flow_dissector,
1258 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1259 target_container);
1260 key_tags->flow_label = ntohl(flow_label);
1261 }
1262 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1263 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1264 break;
1265 }
1266 }
1267
1268 __skb_flow_dissect_ipv6(skb, flow_dissector,
1269 target_container, data, iph);
1270
1271 break;
1272 }
1273 case htons(ETH_P_8021AD):
1274 case htons(ETH_P_8021Q): {
1275 const struct vlan_hdr *vlan = NULL;
1276 struct vlan_hdr _vlan;
1277 __be16 saved_vlan_tpid = proto;
1278
1279 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1280 skb && skb_vlan_tag_present(skb)) {
1281 proto = skb->protocol;
1282 } else {
1283 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1284 data, hlen, &_vlan);
1285 if (!vlan) {
1286 fdret = FLOW_DISSECT_RET_OUT_BAD;
1287 break;
1288 }
1289
1290 proto = vlan->h_vlan_encapsulated_proto;
1291 nhoff += sizeof(*vlan);
1292 }
1293
1294 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1295 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1296 struct flow_dissector_key_num_of_vlans *key_nvs;
1297
1298 key_nvs = skb_flow_dissector_target(flow_dissector,
1299 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1300 target_container);
1301 key_nvs->num_of_vlans++;
1302 }
1303
1304 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1305 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1306 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1307 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1308 } else {
1309 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1310 break;
1311 }
1312
1313 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1314 key_vlan = skb_flow_dissector_target(flow_dissector,
1315 dissector_vlan,
1316 target_container);
1317
1318 if (!vlan) {
1319 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1320 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1321 } else {
1322 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1323 VLAN_VID_MASK;
1324 key_vlan->vlan_priority =
1325 (ntohs(vlan->h_vlan_TCI) &
1326 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1327 }
1328 key_vlan->vlan_tpid = saved_vlan_tpid;
1329 key_vlan->vlan_eth_type = proto;
1330 }
1331
1332 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1333 break;
1334 }
1335 case htons(ETH_P_PPP_SES): {
1336 struct {
1337 struct pppoe_hdr hdr;
1338 __be16 proto;
1339 } *hdr, _hdr;
1340 u16 ppp_proto;
1341
1342 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1343 if (!hdr) {
1344 fdret = FLOW_DISSECT_RET_OUT_BAD;
1345 break;
1346 }
1347
1348 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1349 fdret = FLOW_DISSECT_RET_OUT_BAD;
1350 break;
1351 }
1352
1353 /* least significant bit of the most significant octet
1354 * indicates if protocol field was compressed
1355 */
1356 ppp_proto = ntohs(hdr->proto);
1357 if (ppp_proto & 0x0100) {
1358 ppp_proto = ppp_proto >> 8;
1359 nhoff += PPPOE_SES_HLEN - 1;
1360 } else {
1361 nhoff += PPPOE_SES_HLEN;
1362 }
1363
1364 if (ppp_proto == PPP_IP) {
1365 proto = htons(ETH_P_IP);
1366 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1367 } else if (ppp_proto == PPP_IPV6) {
1368 proto = htons(ETH_P_IPV6);
1369 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1370 } else if (ppp_proto == PPP_MPLS_UC) {
1371 proto = htons(ETH_P_MPLS_UC);
1372 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1373 } else if (ppp_proto == PPP_MPLS_MC) {
1374 proto = htons(ETH_P_MPLS_MC);
1375 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1376 } else if (ppp_proto_is_valid(ppp_proto)) {
1377 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1378 } else {
1379 fdret = FLOW_DISSECT_RET_OUT_BAD;
1380 break;
1381 }
1382
1383 if (dissector_uses_key(flow_dissector,
1384 FLOW_DISSECTOR_KEY_PPPOE)) {
1385 struct flow_dissector_key_pppoe *key_pppoe;
1386
1387 key_pppoe = skb_flow_dissector_target(flow_dissector,
1388 FLOW_DISSECTOR_KEY_PPPOE,
1389 target_container);
1390 key_pppoe->session_id = hdr->hdr.sid;
1391 key_pppoe->ppp_proto = htons(ppp_proto);
1392 key_pppoe->type = htons(ETH_P_PPP_SES);
1393 }
1394 break;
1395 }
1396 case htons(ETH_P_TIPC): {
1397 struct tipc_basic_hdr *hdr, _hdr;
1398
1399 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1400 data, hlen, &_hdr);
1401 if (!hdr) {
1402 fdret = FLOW_DISSECT_RET_OUT_BAD;
1403 break;
1404 }
1405
1406 if (dissector_uses_key(flow_dissector,
1407 FLOW_DISSECTOR_KEY_TIPC)) {
1408 key_addrs = skb_flow_dissector_target(flow_dissector,
1409 FLOW_DISSECTOR_KEY_TIPC,
1410 target_container);
1411 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1412 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1413 }
1414 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1415 break;
1416 }
1417
1418 case htons(ETH_P_MPLS_UC):
1419 case htons(ETH_P_MPLS_MC):
1420 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1421 target_container, data,
1422 nhoff, hlen, mpls_lse,
1423 &mpls_el);
1424 nhoff += sizeof(struct mpls_label);
1425 mpls_lse++;
1426 break;
1427 case htons(ETH_P_FCOE):
1428 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1429 fdret = FLOW_DISSECT_RET_OUT_BAD;
1430 break;
1431 }
1432
1433 nhoff += FCOE_HEADER_LEN;
1434 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1435 break;
1436
1437 case htons(ETH_P_ARP):
1438 case htons(ETH_P_RARP):
1439 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1440 target_container, data,
1441 nhoff, hlen);
1442 break;
1443
1444 case htons(ETH_P_BATMAN):
1445 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1446 &proto, &nhoff, hlen, flags);
1447 break;
1448
1449 case htons(ETH_P_1588): {
1450 struct ptp_header *hdr, _hdr;
1451
1452 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1453 hlen, &_hdr);
1454 if (!hdr) {
1455 fdret = FLOW_DISSECT_RET_OUT_BAD;
1456 break;
1457 }
1458
1459 nhoff += sizeof(struct ptp_header);
1460 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1461 break;
1462 }
1463
1464 case htons(ETH_P_PRP):
1465 case htons(ETH_P_HSR): {
1466 struct hsr_tag *hdr, _hdr;
1467
1468 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1469 &_hdr);
1470 if (!hdr) {
1471 fdret = FLOW_DISSECT_RET_OUT_BAD;
1472 break;
1473 }
1474
1475 proto = hdr->encap_proto;
1476 nhoff += HSR_HLEN;
1477 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1478 break;
1479 }
1480
1481 case htons(ETH_P_CFM):
1482 fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1483 target_container, data,
1484 nhoff, hlen);
1485 break;
1486
1487 default:
1488 fdret = FLOW_DISSECT_RET_OUT_BAD;
1489 break;
1490 }
1491
1492 /* Process result of proto processing */
1493 switch (fdret) {
1494 case FLOW_DISSECT_RET_OUT_GOOD:
1495 goto out_good;
1496 case FLOW_DISSECT_RET_PROTO_AGAIN:
1497 if (skb_flow_dissect_allowed(&num_hdrs))
1498 goto proto_again;
1499 goto out_good;
1500 case FLOW_DISSECT_RET_CONTINUE:
1501 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1502 break;
1503 case FLOW_DISSECT_RET_OUT_BAD:
1504 default:
1505 goto out_bad;
1506 }
1507
1508 ip_proto_again:
1509 fdret = FLOW_DISSECT_RET_CONTINUE;
1510
1511 switch (ip_proto) {
1512 case IPPROTO_GRE:
1513 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1514 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1515 break;
1516 }
1517
1518 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1519 target_container, data,
1520 &proto, &nhoff, &hlen, flags);
1521 break;
1522
1523 case NEXTHDR_HOP:
1524 case NEXTHDR_ROUTING:
1525 case NEXTHDR_DEST: {
1526 u8 _opthdr[2], *opthdr;
1527
1528 if (proto != htons(ETH_P_IPV6))
1529 break;
1530
1531 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1532 data, hlen, &_opthdr);
1533 if (!opthdr) {
1534 fdret = FLOW_DISSECT_RET_OUT_BAD;
1535 break;
1536 }
1537
1538 ip_proto = opthdr[0];
1539 nhoff += (opthdr[1] + 1) << 3;
1540
1541 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1542 break;
1543 }
1544 case NEXTHDR_FRAGMENT: {
1545 struct frag_hdr _fh, *fh;
1546
1547 if (proto != htons(ETH_P_IPV6))
1548 break;
1549
1550 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1551 data, hlen, &_fh);
1552
1553 if (!fh) {
1554 fdret = FLOW_DISSECT_RET_OUT_BAD;
1555 break;
1556 }
1557
1558 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1559
1560 nhoff += sizeof(_fh);
1561 ip_proto = fh->nexthdr;
1562
1563 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1564 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1565 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1566 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1567 break;
1568 }
1569 }
1570
1571 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1572 break;
1573 }
1574 case IPPROTO_IPIP:
1575 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1576 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1577 break;
1578 }
1579
1580 proto = htons(ETH_P_IP);
1581
1582 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1583 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1584 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1585 break;
1586 }
1587
1588 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1589 break;
1590
1591 case IPPROTO_IPV6:
1592 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1593 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1594 break;
1595 }
1596
1597 proto = htons(ETH_P_IPV6);
1598
1599 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1600 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1601 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1602 break;
1603 }
1604
1605 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1606 break;
1607
1608
1609 case IPPROTO_MPLS:
1610 proto = htons(ETH_P_MPLS_UC);
1611 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1612 break;
1613
1614 case IPPROTO_TCP:
1615 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1616 data, nhoff, hlen);
1617 break;
1618
1619 case IPPROTO_ICMP:
1620 case IPPROTO_ICMPV6:
1621 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1622 data, nhoff, hlen);
1623 break;
1624 case IPPROTO_L2TP:
1625 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1626 data, nhoff, hlen);
1627 break;
1628 case IPPROTO_ESP:
1629 __skb_flow_dissect_esp(skb, flow_dissector, target_container,
1630 data, nhoff, hlen);
1631 break;
1632 case IPPROTO_AH:
1633 __skb_flow_dissect_ah(skb, flow_dissector, target_container,
1634 data, nhoff, hlen);
1635 break;
1636 default:
1637 break;
1638 }
1639
1640 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1641 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1642 data, nhoff, ip_proto, hlen);
1643
1644 /* Process result of IP proto processing */
1645 switch (fdret) {
1646 case FLOW_DISSECT_RET_PROTO_AGAIN:
1647 if (skb_flow_dissect_allowed(&num_hdrs))
1648 goto proto_again;
1649 break;
1650 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1651 if (skb_flow_dissect_allowed(&num_hdrs))
1652 goto ip_proto_again;
1653 break;
1654 case FLOW_DISSECT_RET_OUT_GOOD:
1655 case FLOW_DISSECT_RET_CONTINUE:
1656 break;
1657 case FLOW_DISSECT_RET_OUT_BAD:
1658 default:
1659 goto out_bad;
1660 }
1661
1662 out_good:
1663 ret = true;
1664
1665 out:
1666 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1667 key_basic->n_proto = proto;
1668 key_basic->ip_proto = ip_proto;
1669
1670 return ret;
1671
1672 out_bad:
1673 ret = false;
1674 goto out;
1675 }
1676 EXPORT_SYMBOL(__skb_flow_dissect);
1677
1678 static siphash_aligned_key_t hashrnd;
__flow_hash_secret_init(void)1679 static __always_inline void __flow_hash_secret_init(void)
1680 {
1681 net_get_random_once(&hashrnd, sizeof(hashrnd));
1682 }
1683
flow_keys_hash_start(const struct flow_keys * flow)1684 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1685 {
1686 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1687 return &flow->FLOW_KEYS_HASH_START_FIELD;
1688 }
1689
flow_keys_hash_length(const struct flow_keys * flow)1690 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1691 {
1692 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1693
1694 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1695
1696 switch (flow->control.addr_type) {
1697 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1698 diff -= sizeof(flow->addrs.v4addrs);
1699 break;
1700 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1701 diff -= sizeof(flow->addrs.v6addrs);
1702 break;
1703 case FLOW_DISSECTOR_KEY_TIPC:
1704 diff -= sizeof(flow->addrs.tipckey);
1705 break;
1706 }
1707 return sizeof(*flow) - diff;
1708 }
1709
flow_get_u32_src(const struct flow_keys * flow)1710 __be32 flow_get_u32_src(const struct flow_keys *flow)
1711 {
1712 switch (flow->control.addr_type) {
1713 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1714 return flow->addrs.v4addrs.src;
1715 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1716 return (__force __be32)ipv6_addr_hash(
1717 &flow->addrs.v6addrs.src);
1718 case FLOW_DISSECTOR_KEY_TIPC:
1719 return flow->addrs.tipckey.key;
1720 default:
1721 return 0;
1722 }
1723 }
1724 EXPORT_SYMBOL(flow_get_u32_src);
1725
flow_get_u32_dst(const struct flow_keys * flow)1726 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1727 {
1728 switch (flow->control.addr_type) {
1729 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1730 return flow->addrs.v4addrs.dst;
1731 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1732 return (__force __be32)ipv6_addr_hash(
1733 &flow->addrs.v6addrs.dst);
1734 default:
1735 return 0;
1736 }
1737 }
1738 EXPORT_SYMBOL(flow_get_u32_dst);
1739
1740 /* Sort the source and destination IP and the ports,
1741 * to have consistent hash within the two directions
1742 */
__flow_hash_consistentify(struct flow_keys * keys)1743 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1744 {
1745 int addr_diff, i;
1746
1747 switch (keys->control.addr_type) {
1748 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1749 if ((__force u32)keys->addrs.v4addrs.dst <
1750 (__force u32)keys->addrs.v4addrs.src)
1751 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1752
1753 if ((__force u16)keys->ports.dst <
1754 (__force u16)keys->ports.src) {
1755 swap(keys->ports.src, keys->ports.dst);
1756 }
1757 break;
1758 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1759 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1760 &keys->addrs.v6addrs.src,
1761 sizeof(keys->addrs.v6addrs.dst));
1762 if (addr_diff < 0) {
1763 for (i = 0; i < 4; i++)
1764 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1765 keys->addrs.v6addrs.dst.s6_addr32[i]);
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 }
1773 }
1774
__flow_hash_from_keys(struct flow_keys * keys,const siphash_key_t * keyval)1775 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1776 const siphash_key_t *keyval)
1777 {
1778 u32 hash;
1779
1780 __flow_hash_consistentify(keys);
1781
1782 hash = siphash(flow_keys_hash_start(keys),
1783 flow_keys_hash_length(keys), keyval);
1784 if (!hash)
1785 hash = 1;
1786
1787 return hash;
1788 }
1789
flow_hash_from_keys(struct flow_keys * keys)1790 u32 flow_hash_from_keys(struct flow_keys *keys)
1791 {
1792 __flow_hash_secret_init();
1793 return __flow_hash_from_keys(keys, &hashrnd);
1794 }
1795 EXPORT_SYMBOL(flow_hash_from_keys);
1796
___skb_get_hash(const struct sk_buff * skb,struct flow_keys * keys,const siphash_key_t * keyval)1797 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1798 struct flow_keys *keys,
1799 const siphash_key_t *keyval)
1800 {
1801 skb_flow_dissect_flow_keys(skb, keys,
1802 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1803
1804 return __flow_hash_from_keys(keys, keyval);
1805 }
1806
1807 struct _flow_keys_digest_data {
1808 __be16 n_proto;
1809 u8 ip_proto;
1810 u8 padding;
1811 __be32 ports;
1812 __be32 src;
1813 __be32 dst;
1814 };
1815
make_flow_keys_digest(struct flow_keys_digest * digest,const struct flow_keys * flow)1816 void make_flow_keys_digest(struct flow_keys_digest *digest,
1817 const struct flow_keys *flow)
1818 {
1819 struct _flow_keys_digest_data *data =
1820 (struct _flow_keys_digest_data *)digest;
1821
1822 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1823
1824 memset(digest, 0, sizeof(*digest));
1825
1826 data->n_proto = flow->basic.n_proto;
1827 data->ip_proto = flow->basic.ip_proto;
1828 data->ports = flow->ports.ports;
1829 data->src = flow->addrs.v4addrs.src;
1830 data->dst = flow->addrs.v4addrs.dst;
1831 }
1832 EXPORT_SYMBOL(make_flow_keys_digest);
1833
1834 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1835
__skb_get_hash_symmetric(const struct sk_buff * skb)1836 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1837 {
1838 struct flow_keys keys;
1839
1840 __flow_hash_secret_init();
1841
1842 memset(&keys, 0, sizeof(keys));
1843 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1844 &keys, NULL, 0, 0, 0, 0);
1845
1846 return __flow_hash_from_keys(&keys, &hashrnd);
1847 }
1848 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1849
1850 /**
1851 * __skb_get_hash: calculate a flow hash
1852 * @skb: sk_buff to calculate flow hash from
1853 *
1854 * This function calculates a flow hash based on src/dst addresses
1855 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1856 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1857 * if hash is a canonical 4-tuple hash over transport ports.
1858 */
__skb_get_hash(struct sk_buff * skb)1859 void __skb_get_hash(struct sk_buff *skb)
1860 {
1861 struct flow_keys keys;
1862 u32 hash;
1863
1864 __flow_hash_secret_init();
1865
1866 hash = ___skb_get_hash(skb, &keys, &hashrnd);
1867
1868 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1869 }
1870 EXPORT_SYMBOL(__skb_get_hash);
1871
skb_get_hash_perturb(const struct sk_buff * skb,const siphash_key_t * perturb)1872 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1873 const siphash_key_t *perturb)
1874 {
1875 struct flow_keys keys;
1876
1877 return ___skb_get_hash(skb, &keys, perturb);
1878 }
1879 EXPORT_SYMBOL(skb_get_hash_perturb);
1880
__skb_get_poff(const struct sk_buff * skb,const void * data,const struct flow_keys_basic * keys,int hlen)1881 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1882 const struct flow_keys_basic *keys, int hlen)
1883 {
1884 u32 poff = keys->control.thoff;
1885
1886 /* skip L4 headers for fragments after the first */
1887 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1888 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1889 return poff;
1890
1891 switch (keys->basic.ip_proto) {
1892 case IPPROTO_TCP: {
1893 /* access doff as u8 to avoid unaligned access */
1894 const u8 *doff;
1895 u8 _doff;
1896
1897 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1898 data, hlen, &_doff);
1899 if (!doff)
1900 return poff;
1901
1902 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1903 break;
1904 }
1905 case IPPROTO_UDP:
1906 case IPPROTO_UDPLITE:
1907 poff += sizeof(struct udphdr);
1908 break;
1909 /* For the rest, we do not really care about header
1910 * extensions at this point for now.
1911 */
1912 case IPPROTO_ICMP:
1913 poff += sizeof(struct icmphdr);
1914 break;
1915 case IPPROTO_ICMPV6:
1916 poff += sizeof(struct icmp6hdr);
1917 break;
1918 case IPPROTO_IGMP:
1919 poff += sizeof(struct igmphdr);
1920 break;
1921 case IPPROTO_DCCP:
1922 poff += sizeof(struct dccp_hdr);
1923 break;
1924 case IPPROTO_SCTP:
1925 poff += sizeof(struct sctphdr);
1926 break;
1927 }
1928
1929 return poff;
1930 }
1931
1932 /**
1933 * skb_get_poff - get the offset to the payload
1934 * @skb: sk_buff to get the payload offset from
1935 *
1936 * The function will get the offset to the payload as far as it could
1937 * be dissected. The main user is currently BPF, so that we can dynamically
1938 * truncate packets without needing to push actual payload to the user
1939 * space and can analyze headers only, instead.
1940 */
skb_get_poff(const struct sk_buff * skb)1941 u32 skb_get_poff(const struct sk_buff *skb)
1942 {
1943 struct flow_keys_basic keys;
1944
1945 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1946 NULL, 0, 0, 0, 0))
1947 return 0;
1948
1949 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1950 }
1951
__get_hash_from_flowi6(const struct flowi6 * fl6,struct flow_keys * keys)1952 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1953 {
1954 memset(keys, 0, sizeof(*keys));
1955
1956 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1957 sizeof(keys->addrs.v6addrs.src));
1958 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1959 sizeof(keys->addrs.v6addrs.dst));
1960 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1961 keys->ports.src = fl6->fl6_sport;
1962 keys->ports.dst = fl6->fl6_dport;
1963 keys->keyid.keyid = fl6->fl6_gre_key;
1964 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1965 keys->basic.ip_proto = fl6->flowi6_proto;
1966
1967 return flow_hash_from_keys(keys);
1968 }
1969 EXPORT_SYMBOL(__get_hash_from_flowi6);
1970
1971 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1972 {
1973 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1974 .offset = offsetof(struct flow_keys, control),
1975 },
1976 {
1977 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1978 .offset = offsetof(struct flow_keys, basic),
1979 },
1980 {
1981 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1982 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1983 },
1984 {
1985 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1986 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1987 },
1988 {
1989 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1990 .offset = offsetof(struct flow_keys, addrs.tipckey),
1991 },
1992 {
1993 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1994 .offset = offsetof(struct flow_keys, ports),
1995 },
1996 {
1997 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1998 .offset = offsetof(struct flow_keys, vlan),
1999 },
2000 {
2001 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2002 .offset = offsetof(struct flow_keys, tags),
2003 },
2004 {
2005 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2006 .offset = offsetof(struct flow_keys, keyid),
2007 },
2008 };
2009
2010 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2011 {
2012 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2013 .offset = offsetof(struct flow_keys, control),
2014 },
2015 {
2016 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2017 .offset = offsetof(struct flow_keys, basic),
2018 },
2019 {
2020 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2021 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2022 },
2023 {
2024 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2025 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2026 },
2027 {
2028 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2029 .offset = offsetof(struct flow_keys, ports),
2030 },
2031 };
2032
2033 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2034 {
2035 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2036 .offset = offsetof(struct flow_keys, control),
2037 },
2038 {
2039 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2040 .offset = offsetof(struct flow_keys, basic),
2041 },
2042 };
2043
2044 struct flow_dissector flow_keys_dissector __read_mostly;
2045 EXPORT_SYMBOL(flow_keys_dissector);
2046
2047 struct flow_dissector flow_keys_basic_dissector __read_mostly;
2048 EXPORT_SYMBOL(flow_keys_basic_dissector);
2049
init_default_flow_dissectors(void)2050 static int __init init_default_flow_dissectors(void)
2051 {
2052 skb_flow_dissector_init(&flow_keys_dissector,
2053 flow_keys_dissector_keys,
2054 ARRAY_SIZE(flow_keys_dissector_keys));
2055 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2056 flow_keys_dissector_symmetric_keys,
2057 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2058 skb_flow_dissector_init(&flow_keys_basic_dissector,
2059 flow_keys_basic_dissector_keys,
2060 ARRAY_SIZE(flow_keys_basic_dissector_keys));
2061 return 0;
2062 }
2063 core_initcall(init_default_flow_dissectors);
2064