1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/kernel.h>
3 #include <linux/skbuff.h>
4 #include <linux/export.h>
5 #include <linux/ip.h>
6 #include <linux/ipv6.h>
7 #include <linux/if_vlan.h>
8 #include <linux/filter.h>
9 #include <net/dsa.h>
10 #include <net/dst_metadata.h>
11 #include <net/ip.h>
12 #include <net/ipv6.h>
13 #include <net/gre.h>
14 #include <net/pptp.h>
15 #include <net/tipc.h>
16 #include <linux/igmp.h>
17 #include <linux/icmp.h>
18 #include <linux/sctp.h>
19 #include <linux/dccp.h>
20 #include <linux/if_tunnel.h>
21 #include <linux/if_pppox.h>
22 #include <linux/ppp_defs.h>
23 #include <linux/stddef.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_hsr.h>
26 #include <linux/mpls.h>
27 #include <linux/tcp.h>
28 #include <linux/ptp_classify.h>
29 #include <net/flow_dissector.h>
30 #include <net/pkt_cls.h>
31 #include <scsi/fc/fc_fcoe.h>
32 #include <uapi/linux/batadv_packet.h>
33 #include <linux/bpf.h>
34 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
35 #include <net/netfilter/nf_conntrack_core.h>
36 #include <net/netfilter/nf_conntrack_labels.h>
37 #endif
38 #include <linux/bpf-netns.h>
39
dissector_set_key(struct flow_dissector * flow_dissector,enum flow_dissector_key_id key_id)40 static void dissector_set_key(struct flow_dissector *flow_dissector,
41 enum flow_dissector_key_id key_id)
42 {
43 flow_dissector->used_keys |= (1ULL << key_id);
44 }
45
skb_flow_dissector_init(struct flow_dissector * flow_dissector,const struct flow_dissector_key * key,unsigned int key_count)46 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
47 const struct flow_dissector_key *key,
48 unsigned int key_count)
49 {
50 unsigned int i;
51
52 memset(flow_dissector, 0, sizeof(*flow_dissector));
53
54 for (i = 0; i < key_count; i++, key++) {
55 /* User should make sure that every key target offset is within
56 * boundaries of unsigned short.
57 */
58 BUG_ON(key->offset > USHRT_MAX);
59 BUG_ON(dissector_uses_key(flow_dissector,
60 key->key_id));
61
62 dissector_set_key(flow_dissector, key->key_id);
63 flow_dissector->offset[key->key_id] = key->offset;
64 }
65
66 /* Ensure that the dissector always includes control and basic key.
67 * That way we are able to avoid handling lack of these in fast path.
68 */
69 BUG_ON(!dissector_uses_key(flow_dissector,
70 FLOW_DISSECTOR_KEY_CONTROL));
71 BUG_ON(!dissector_uses_key(flow_dissector,
72 FLOW_DISSECTOR_KEY_BASIC));
73 }
74 EXPORT_SYMBOL(skb_flow_dissector_init);
75
76 #ifdef CONFIG_BPF_SYSCALL
flow_dissector_bpf_prog_attach_check(struct net * net,struct bpf_prog * prog)77 int flow_dissector_bpf_prog_attach_check(struct net *net,
78 struct bpf_prog *prog)
79 {
80 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
81
82 if (net == &init_net) {
83 /* BPF flow dissector in the root namespace overrides
84 * any per-net-namespace one. When attaching to root,
85 * make sure we don't have any BPF program attached
86 * to the non-root namespaces.
87 */
88 struct net *ns;
89
90 for_each_net(ns) {
91 if (ns == &init_net)
92 continue;
93 if (rcu_access_pointer(ns->bpf.run_array[type]))
94 return -EEXIST;
95 }
96 } else {
97 /* Make sure root flow dissector is not attached
98 * when attaching to the non-root namespace.
99 */
100 if (rcu_access_pointer(init_net.bpf.run_array[type]))
101 return -EEXIST;
102 }
103
104 return 0;
105 }
106 #endif /* CONFIG_BPF_SYSCALL */
107
108 /**
109 * skb_flow_get_ports - extract the upper layer ports and return them
110 * @skb: sk_buff to extract the ports from
111 * @thoff: transport header offset
112 * @ip_proto: protocol for which to get port offset
113 * @data: raw buffer pointer to the packet, if NULL use skb->data
114 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
115 *
116 * The function will try to retrieve the ports at offset thoff + poff where poff
117 * is the protocol port offset returned from proto_ports_offset
118 */
skb_flow_get_ports(const struct sk_buff * skb,int thoff,u8 ip_proto,const void * data,int hlen)119 __be32 skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
120 const void *data, int hlen)
121 {
122 int poff = proto_ports_offset(ip_proto);
123
124 if (!data) {
125 data = skb->data;
126 hlen = skb_headlen(skb);
127 }
128
129 if (poff >= 0) {
130 __be32 *ports, _ports;
131
132 ports = __skb_header_pointer(skb, thoff + poff,
133 sizeof(_ports), data, hlen, &_ports);
134 if (ports)
135 return *ports;
136 }
137
138 return 0;
139 }
140 EXPORT_SYMBOL(skb_flow_get_ports);
141
icmp_has_id(u8 type)142 static bool icmp_has_id(u8 type)
143 {
144 switch (type) {
145 case ICMP_ECHO:
146 case ICMP_ECHOREPLY:
147 case ICMP_TIMESTAMP:
148 case ICMP_TIMESTAMPREPLY:
149 case ICMPV6_ECHO_REQUEST:
150 case ICMPV6_ECHO_REPLY:
151 return true;
152 }
153
154 return false;
155 }
156
157 /**
158 * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
159 * @skb: sk_buff to extract from
160 * @key_icmp: struct flow_dissector_key_icmp to fill
161 * @data: raw buffer pointer to the packet
162 * @thoff: offset to extract at
163 * @hlen: packet header length
164 */
skb_flow_get_icmp_tci(const struct sk_buff * skb,struct flow_dissector_key_icmp * key_icmp,const void * data,int thoff,int hlen)165 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
166 struct flow_dissector_key_icmp *key_icmp,
167 const void *data, int thoff, int hlen)
168 {
169 struct icmphdr *ih, _ih;
170
171 ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
172 if (!ih)
173 return;
174
175 key_icmp->type = ih->type;
176 key_icmp->code = ih->code;
177
178 /* As we use 0 to signal that the Id field is not present,
179 * avoid confusion with packets without such field
180 */
181 if (icmp_has_id(ih->type))
182 key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
183 else
184 key_icmp->id = 0;
185 }
186 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
187
188 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
189 * using skb_flow_get_icmp_tci().
190 */
__skb_flow_dissect_icmp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int thoff,int hlen)191 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
192 struct flow_dissector *flow_dissector,
193 void *target_container, const void *data,
194 int thoff, int hlen)
195 {
196 struct flow_dissector_key_icmp *key_icmp;
197
198 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
199 return;
200
201 key_icmp = skb_flow_dissector_target(flow_dissector,
202 FLOW_DISSECTOR_KEY_ICMP,
203 target_container);
204
205 skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
206 }
207
__skb_flow_dissect_ah(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)208 static void __skb_flow_dissect_ah(const struct sk_buff *skb,
209 struct flow_dissector *flow_dissector,
210 void *target_container, const void *data,
211 int nhoff, int hlen)
212 {
213 struct flow_dissector_key_ipsec *key_ah;
214 struct ip_auth_hdr _hdr, *hdr;
215
216 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
217 return;
218
219 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
220 if (!hdr)
221 return;
222
223 key_ah = skb_flow_dissector_target(flow_dissector,
224 FLOW_DISSECTOR_KEY_IPSEC,
225 target_container);
226
227 key_ah->spi = hdr->spi;
228 }
229
__skb_flow_dissect_esp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)230 static void __skb_flow_dissect_esp(const struct sk_buff *skb,
231 struct flow_dissector *flow_dissector,
232 void *target_container, const void *data,
233 int nhoff, int hlen)
234 {
235 struct flow_dissector_key_ipsec *key_esp;
236 struct ip_esp_hdr _hdr, *hdr;
237
238 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPSEC))
239 return;
240
241 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
242 if (!hdr)
243 return;
244
245 key_esp = skb_flow_dissector_target(flow_dissector,
246 FLOW_DISSECTOR_KEY_IPSEC,
247 target_container);
248
249 key_esp->spi = hdr->spi;
250 }
251
__skb_flow_dissect_l2tpv3(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)252 static void __skb_flow_dissect_l2tpv3(const struct sk_buff *skb,
253 struct flow_dissector *flow_dissector,
254 void *target_container, const void *data,
255 int nhoff, int hlen)
256 {
257 struct flow_dissector_key_l2tpv3 *key_l2tpv3;
258 struct {
259 __be32 session_id;
260 } *hdr, _hdr;
261
262 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3))
263 return;
264
265 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
266 if (!hdr)
267 return;
268
269 key_l2tpv3 = skb_flow_dissector_target(flow_dissector,
270 FLOW_DISSECTOR_KEY_L2TPV3,
271 target_container);
272
273 key_l2tpv3->session_id = hdr->session_id;
274 }
275
skb_flow_dissect_meta(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)276 void skb_flow_dissect_meta(const struct sk_buff *skb,
277 struct flow_dissector *flow_dissector,
278 void *target_container)
279 {
280 struct flow_dissector_key_meta *meta;
281
282 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
283 return;
284
285 meta = skb_flow_dissector_target(flow_dissector,
286 FLOW_DISSECTOR_KEY_META,
287 target_container);
288 meta->ingress_ifindex = skb->skb_iif;
289 #if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
290 if (tc_skb_ext_tc_enabled()) {
291 struct tc_skb_ext *ext;
292
293 ext = skb_ext_find(skb, TC_SKB_EXT);
294 if (ext)
295 meta->l2_miss = ext->l2_miss;
296 }
297 #endif
298 }
299 EXPORT_SYMBOL(skb_flow_dissect_meta);
300
301 static void
skb_flow_dissect_set_enc_control(enum flow_dissector_key_id type,u32 ctrl_flags,struct flow_dissector * flow_dissector,void * target_container)302 skb_flow_dissect_set_enc_control(enum flow_dissector_key_id type,
303 u32 ctrl_flags,
304 struct flow_dissector *flow_dissector,
305 void *target_container)
306 {
307 struct flow_dissector_key_control *ctrl;
308
309 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
310 return;
311
312 ctrl = skb_flow_dissector_target(flow_dissector,
313 FLOW_DISSECTOR_KEY_ENC_CONTROL,
314 target_container);
315 ctrl->addr_type = type;
316 ctrl->flags = ctrl_flags;
317 }
318
319 void
skb_flow_dissect_ct(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,u16 * ctinfo_map,size_t mapsize,bool post_ct,u16 zone)320 skb_flow_dissect_ct(const struct sk_buff *skb,
321 struct flow_dissector *flow_dissector,
322 void *target_container, u16 *ctinfo_map,
323 size_t mapsize, bool post_ct, u16 zone)
324 {
325 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
326 struct flow_dissector_key_ct *key;
327 enum ip_conntrack_info ctinfo;
328 struct nf_conn_labels *cl;
329 struct nf_conn *ct;
330
331 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
332 return;
333
334 ct = nf_ct_get(skb, &ctinfo);
335 if (!ct && !post_ct)
336 return;
337
338 key = skb_flow_dissector_target(flow_dissector,
339 FLOW_DISSECTOR_KEY_CT,
340 target_container);
341
342 if (!ct) {
343 key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
344 TCA_FLOWER_KEY_CT_FLAGS_INVALID;
345 key->ct_zone = zone;
346 return;
347 }
348
349 if (ctinfo < mapsize)
350 key->ct_state = ctinfo_map[ctinfo];
351 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
352 key->ct_zone = ct->zone.id;
353 #endif
354 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
355 key->ct_mark = READ_ONCE(ct->mark);
356 #endif
357
358 cl = nf_ct_labels_find(ct);
359 if (cl)
360 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
361 #endif /* CONFIG_NF_CONNTRACK */
362 }
363 EXPORT_SYMBOL(skb_flow_dissect_ct);
364
365 void
skb_flow_dissect_tunnel_info(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)366 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
367 struct flow_dissector *flow_dissector,
368 void *target_container)
369 {
370 struct ip_tunnel_info *info;
371 struct ip_tunnel_key *key;
372 u32 ctrl_flags = 0;
373
374 /* A quick check to see if there might be something to do. */
375 if (!dissector_uses_key(flow_dissector,
376 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
377 !dissector_uses_key(flow_dissector,
378 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
379 !dissector_uses_key(flow_dissector,
380 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
381 !dissector_uses_key(flow_dissector,
382 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
383 !dissector_uses_key(flow_dissector,
384 FLOW_DISSECTOR_KEY_ENC_PORTS) &&
385 !dissector_uses_key(flow_dissector,
386 FLOW_DISSECTOR_KEY_ENC_IP) &&
387 !dissector_uses_key(flow_dissector,
388 FLOW_DISSECTOR_KEY_ENC_OPTS))
389 return;
390
391 info = skb_tunnel_info(skb);
392 if (!info)
393 return;
394
395 key = &info->key;
396
397 if (test_bit(IP_TUNNEL_CSUM_BIT, key->tun_flags))
398 ctrl_flags |= FLOW_DIS_F_TUNNEL_CSUM;
399 if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT, key->tun_flags))
400 ctrl_flags |= FLOW_DIS_F_TUNNEL_DONT_FRAGMENT;
401 if (test_bit(IP_TUNNEL_OAM_BIT, key->tun_flags))
402 ctrl_flags |= FLOW_DIS_F_TUNNEL_OAM;
403 if (test_bit(IP_TUNNEL_CRIT_OPT_BIT, key->tun_flags))
404 ctrl_flags |= FLOW_DIS_F_TUNNEL_CRIT_OPT;
405
406 switch (ip_tunnel_info_af(info)) {
407 case AF_INET:
408 skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
409 ctrl_flags, flow_dissector,
410 target_container);
411 if (dissector_uses_key(flow_dissector,
412 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
413 struct flow_dissector_key_ipv4_addrs *ipv4;
414
415 ipv4 = skb_flow_dissector_target(flow_dissector,
416 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
417 target_container);
418 ipv4->src = key->u.ipv4.src;
419 ipv4->dst = key->u.ipv4.dst;
420 }
421 break;
422 case AF_INET6:
423 skb_flow_dissect_set_enc_control(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
424 ctrl_flags, flow_dissector,
425 target_container);
426 if (dissector_uses_key(flow_dissector,
427 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
428 struct flow_dissector_key_ipv6_addrs *ipv6;
429
430 ipv6 = skb_flow_dissector_target(flow_dissector,
431 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
432 target_container);
433 ipv6->src = key->u.ipv6.src;
434 ipv6->dst = key->u.ipv6.dst;
435 }
436 break;
437 default:
438 skb_flow_dissect_set_enc_control(0, ctrl_flags, flow_dissector,
439 target_container);
440 break;
441 }
442
443 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
444 struct flow_dissector_key_keyid *keyid;
445
446 keyid = skb_flow_dissector_target(flow_dissector,
447 FLOW_DISSECTOR_KEY_ENC_KEYID,
448 target_container);
449 keyid->keyid = tunnel_id_to_key32(key->tun_id);
450 }
451
452 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
453 struct flow_dissector_key_ports *tp;
454
455 tp = skb_flow_dissector_target(flow_dissector,
456 FLOW_DISSECTOR_KEY_ENC_PORTS,
457 target_container);
458 tp->src = key->tp_src;
459 tp->dst = key->tp_dst;
460 }
461
462 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
463 struct flow_dissector_key_ip *ip;
464
465 ip = skb_flow_dissector_target(flow_dissector,
466 FLOW_DISSECTOR_KEY_ENC_IP,
467 target_container);
468 ip->tos = key->tos;
469 ip->ttl = key->ttl;
470 }
471
472 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
473 struct flow_dissector_key_enc_opts *enc_opt;
474 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
475 u32 val;
476
477 enc_opt = skb_flow_dissector_target(flow_dissector,
478 FLOW_DISSECTOR_KEY_ENC_OPTS,
479 target_container);
480
481 if (!info->options_len)
482 return;
483
484 enc_opt->len = info->options_len;
485 ip_tunnel_info_opts_get(enc_opt->data, info);
486
487 ip_tunnel_set_options_present(flags);
488 ip_tunnel_flags_and(flags, info->key.tun_flags, flags);
489
490 val = find_next_bit(flags, __IP_TUNNEL_FLAG_NUM,
491 IP_TUNNEL_GENEVE_OPT_BIT);
492 enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val : 0;
493 }
494 }
495 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
496
skb_flow_dissect_hash(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)497 void skb_flow_dissect_hash(const struct sk_buff *skb,
498 struct flow_dissector *flow_dissector,
499 void *target_container)
500 {
501 struct flow_dissector_key_hash *key;
502
503 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
504 return;
505
506 key = skb_flow_dissector_target(flow_dissector,
507 FLOW_DISSECTOR_KEY_HASH,
508 target_container);
509
510 key->hash = skb_get_hash_raw(skb);
511 }
512 EXPORT_SYMBOL(skb_flow_dissect_hash);
513
514 static enum flow_dissect_ret
__skb_flow_dissect_mpls(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen,int lse_index,bool * entropy_label)515 __skb_flow_dissect_mpls(const struct sk_buff *skb,
516 struct flow_dissector *flow_dissector,
517 void *target_container, const void *data, int nhoff,
518 int hlen, int lse_index, bool *entropy_label)
519 {
520 struct mpls_label *hdr, _hdr;
521 u32 entry, label, bos;
522
523 if (!dissector_uses_key(flow_dissector,
524 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
525 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
526 return FLOW_DISSECT_RET_OUT_GOOD;
527
528 if (lse_index >= FLOW_DIS_MPLS_MAX)
529 return FLOW_DISSECT_RET_OUT_GOOD;
530
531 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
532 hlen, &_hdr);
533 if (!hdr)
534 return FLOW_DISSECT_RET_OUT_BAD;
535
536 entry = ntohl(hdr->entry);
537 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
538 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
539
540 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
541 struct flow_dissector_key_mpls *key_mpls;
542 struct flow_dissector_mpls_lse *lse;
543
544 key_mpls = skb_flow_dissector_target(flow_dissector,
545 FLOW_DISSECTOR_KEY_MPLS,
546 target_container);
547 lse = &key_mpls->ls[lse_index];
548
549 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
550 lse->mpls_bos = bos;
551 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
552 lse->mpls_label = label;
553 dissector_set_mpls_lse(key_mpls, lse_index);
554 }
555
556 if (*entropy_label &&
557 dissector_uses_key(flow_dissector,
558 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
559 struct flow_dissector_key_keyid *key_keyid;
560
561 key_keyid = skb_flow_dissector_target(flow_dissector,
562 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
563 target_container);
564 key_keyid->keyid = cpu_to_be32(label);
565 }
566
567 *entropy_label = label == MPLS_LABEL_ENTROPY;
568
569 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
570 }
571
572 static enum flow_dissect_ret
__skb_flow_dissect_arp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)573 __skb_flow_dissect_arp(const struct sk_buff *skb,
574 struct flow_dissector *flow_dissector,
575 void *target_container, const void *data,
576 int nhoff, int hlen)
577 {
578 struct flow_dissector_key_arp *key_arp;
579 struct {
580 unsigned char ar_sha[ETH_ALEN];
581 unsigned char ar_sip[4];
582 unsigned char ar_tha[ETH_ALEN];
583 unsigned char ar_tip[4];
584 } *arp_eth, _arp_eth;
585 const struct arphdr *arp;
586 struct arphdr _arp;
587
588 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
589 return FLOW_DISSECT_RET_OUT_GOOD;
590
591 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
592 hlen, &_arp);
593 if (!arp)
594 return FLOW_DISSECT_RET_OUT_BAD;
595
596 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
597 arp->ar_pro != htons(ETH_P_IP) ||
598 arp->ar_hln != ETH_ALEN ||
599 arp->ar_pln != 4 ||
600 (arp->ar_op != htons(ARPOP_REPLY) &&
601 arp->ar_op != htons(ARPOP_REQUEST)))
602 return FLOW_DISSECT_RET_OUT_BAD;
603
604 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
605 sizeof(_arp_eth), data,
606 hlen, &_arp_eth);
607 if (!arp_eth)
608 return FLOW_DISSECT_RET_OUT_BAD;
609
610 key_arp = skb_flow_dissector_target(flow_dissector,
611 FLOW_DISSECTOR_KEY_ARP,
612 target_container);
613
614 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
615 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
616
617 /* Only store the lower byte of the opcode;
618 * this covers ARPOP_REPLY and ARPOP_REQUEST.
619 */
620 key_arp->op = ntohs(arp->ar_op) & 0xff;
621
622 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
623 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
624
625 return FLOW_DISSECT_RET_OUT_GOOD;
626 }
627
628 static enum flow_dissect_ret
__skb_flow_dissect_cfm(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,int hlen)629 __skb_flow_dissect_cfm(const struct sk_buff *skb,
630 struct flow_dissector *flow_dissector,
631 void *target_container, const void *data,
632 int nhoff, int hlen)
633 {
634 struct flow_dissector_key_cfm *key, *hdr, _hdr;
635
636 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM))
637 return FLOW_DISSECT_RET_OUT_GOOD;
638
639 hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr);
640 if (!hdr)
641 return FLOW_DISSECT_RET_OUT_BAD;
642
643 key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM,
644 target_container);
645
646 key->mdl_ver = hdr->mdl_ver;
647 key->opcode = hdr->opcode;
648
649 return FLOW_DISSECT_RET_OUT_GOOD;
650 }
651
652 static enum flow_dissect_ret
__skb_flow_dissect_gre(const struct sk_buff * skb,struct flow_dissector_key_control * key_control,struct flow_dissector * flow_dissector,void * target_container,const void * data,__be16 * p_proto,int * p_nhoff,int * p_hlen,unsigned int flags)653 __skb_flow_dissect_gre(const struct sk_buff *skb,
654 struct flow_dissector_key_control *key_control,
655 struct flow_dissector *flow_dissector,
656 void *target_container, const void *data,
657 __be16 *p_proto, int *p_nhoff, int *p_hlen,
658 unsigned int flags)
659 {
660 struct flow_dissector_key_keyid *key_keyid;
661 struct gre_base_hdr *hdr, _hdr;
662 int offset = 0;
663 u16 gre_ver;
664
665 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
666 data, *p_hlen, &_hdr);
667 if (!hdr)
668 return FLOW_DISSECT_RET_OUT_BAD;
669
670 /* Only look inside GRE without routing */
671 if (hdr->flags & GRE_ROUTING)
672 return FLOW_DISSECT_RET_OUT_GOOD;
673
674 /* Only look inside GRE for version 0 and 1 */
675 gre_ver = ntohs(hdr->flags & GRE_VERSION);
676 if (gre_ver > 1)
677 return FLOW_DISSECT_RET_OUT_GOOD;
678
679 *p_proto = hdr->protocol;
680 if (gre_ver) {
681 /* Version1 must be PPTP, and check the flags */
682 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
683 return FLOW_DISSECT_RET_OUT_GOOD;
684 }
685
686 offset += sizeof(struct gre_base_hdr);
687
688 if (hdr->flags & GRE_CSUM)
689 offset += sizeof_field(struct gre_full_hdr, csum) +
690 sizeof_field(struct gre_full_hdr, reserved1);
691
692 if (hdr->flags & GRE_KEY) {
693 const __be32 *keyid;
694 __be32 _keyid;
695
696 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
697 sizeof(_keyid),
698 data, *p_hlen, &_keyid);
699 if (!keyid)
700 return FLOW_DISSECT_RET_OUT_BAD;
701
702 if (dissector_uses_key(flow_dissector,
703 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
704 key_keyid = skb_flow_dissector_target(flow_dissector,
705 FLOW_DISSECTOR_KEY_GRE_KEYID,
706 target_container);
707 if (gre_ver == 0)
708 key_keyid->keyid = *keyid;
709 else
710 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
711 }
712 offset += sizeof_field(struct gre_full_hdr, key);
713 }
714
715 if (hdr->flags & GRE_SEQ)
716 offset += sizeof_field(struct pptp_gre_header, seq);
717
718 if (gre_ver == 0) {
719 if (*p_proto == htons(ETH_P_TEB)) {
720 const struct ethhdr *eth;
721 struct ethhdr _eth;
722
723 eth = __skb_header_pointer(skb, *p_nhoff + offset,
724 sizeof(_eth),
725 data, *p_hlen, &_eth);
726 if (!eth)
727 return FLOW_DISSECT_RET_OUT_BAD;
728 *p_proto = eth->h_proto;
729 offset += sizeof(*eth);
730
731 /* Cap headers that we access via pointers at the
732 * end of the Ethernet header as our maximum alignment
733 * at that point is only 2 bytes.
734 */
735 if (NET_IP_ALIGN)
736 *p_hlen = *p_nhoff + offset;
737 }
738 } else { /* version 1, must be PPTP */
739 u8 _ppp_hdr[PPP_HDRLEN];
740 u8 *ppp_hdr;
741
742 if (hdr->flags & GRE_ACK)
743 offset += sizeof_field(struct pptp_gre_header, ack);
744
745 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
746 sizeof(_ppp_hdr),
747 data, *p_hlen, _ppp_hdr);
748 if (!ppp_hdr)
749 return FLOW_DISSECT_RET_OUT_BAD;
750
751 switch (PPP_PROTOCOL(ppp_hdr)) {
752 case PPP_IP:
753 *p_proto = htons(ETH_P_IP);
754 break;
755 case PPP_IPV6:
756 *p_proto = htons(ETH_P_IPV6);
757 break;
758 default:
759 /* Could probably catch some more like MPLS */
760 break;
761 }
762
763 offset += PPP_HDRLEN;
764 }
765
766 *p_nhoff += offset;
767 key_control->flags |= FLOW_DIS_ENCAPSULATION;
768 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
769 return FLOW_DISSECT_RET_OUT_GOOD;
770
771 return FLOW_DISSECT_RET_PROTO_AGAIN;
772 }
773
774 /**
775 * __skb_flow_dissect_batadv() - dissect batman-adv header
776 * @skb: sk_buff to with the batman-adv header
777 * @key_control: flow dissectors control key
778 * @data: raw buffer pointer to the packet, if NULL use skb->data
779 * @p_proto: pointer used to update the protocol to process next
780 * @p_nhoff: pointer used to update inner network header offset
781 * @hlen: packet header length
782 * @flags: any combination of FLOW_DISSECTOR_F_*
783 *
784 * ETH_P_BATMAN packets are tried to be dissected. Only
785 * &struct batadv_unicast packets are actually processed because they contain an
786 * inner ethernet header and are usually followed by actual network header. This
787 * allows the flow dissector to continue processing the packet.
788 *
789 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
790 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
791 * otherwise FLOW_DISSECT_RET_OUT_BAD
792 */
793 static enum flow_dissect_ret
__skb_flow_dissect_batadv(const struct sk_buff * skb,struct flow_dissector_key_control * key_control,const void * data,__be16 * p_proto,int * p_nhoff,int hlen,unsigned int flags)794 __skb_flow_dissect_batadv(const struct sk_buff *skb,
795 struct flow_dissector_key_control *key_control,
796 const void *data, __be16 *p_proto, int *p_nhoff,
797 int hlen, unsigned int flags)
798 {
799 struct {
800 struct batadv_unicast_packet batadv_unicast;
801 struct ethhdr eth;
802 } *hdr, _hdr;
803
804 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
805 &_hdr);
806 if (!hdr)
807 return FLOW_DISSECT_RET_OUT_BAD;
808
809 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
810 return FLOW_DISSECT_RET_OUT_BAD;
811
812 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
813 return FLOW_DISSECT_RET_OUT_BAD;
814
815 *p_proto = hdr->eth.h_proto;
816 *p_nhoff += sizeof(*hdr);
817
818 key_control->flags |= FLOW_DIS_ENCAPSULATION;
819 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
820 return FLOW_DISSECT_RET_OUT_GOOD;
821
822 return FLOW_DISSECT_RET_PROTO_AGAIN;
823 }
824
825 static void
__skb_flow_dissect_tcp(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int thoff,int hlen)826 __skb_flow_dissect_tcp(const struct sk_buff *skb,
827 struct flow_dissector *flow_dissector,
828 void *target_container, const void *data,
829 int thoff, int hlen)
830 {
831 struct flow_dissector_key_tcp *key_tcp;
832 struct tcphdr *th, _th;
833
834 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
835 return;
836
837 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
838 if (!th)
839 return;
840
841 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
842 return;
843
844 key_tcp = skb_flow_dissector_target(flow_dissector,
845 FLOW_DISSECTOR_KEY_TCP,
846 target_container);
847 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
848 }
849
850 static void
__skb_flow_dissect_ports(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container,const void * data,int nhoff,u8 ip_proto,int hlen)851 __skb_flow_dissect_ports(const struct sk_buff *skb,
852 struct flow_dissector *flow_dissector,
853 void *target_container, const void *data,
854 int nhoff, u8 ip_proto, int hlen)
855 {
856 struct flow_dissector_key_ports_range *key_ports_range = NULL;
857 struct flow_dissector_key_ports *key_ports = NULL;
858 __be32 ports;
859
860 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
861 key_ports = skb_flow_dissector_target(flow_dissector,
862 FLOW_DISSECTOR_KEY_PORTS,
863 target_container);
864
865 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS_RANGE))
866 key_ports_range = skb_flow_dissector_target(flow_dissector,
867 FLOW_DISSECTOR_KEY_PORTS_RANGE,
868 target_container);
869
870 if (!key_ports && !key_ports_range)
871 return;
872
873 ports = skb_flow_get_ports(skb, nhoff, ip_proto, data, hlen);
874
875 if (key_ports)
876 key_ports->ports = ports;
877
878 if (key_ports_range)
879 key_ports_range->tp.ports = ports;
880 }
881
882 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)883 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
884 struct flow_dissector *flow_dissector,
885 void *target_container, const void *data,
886 const struct iphdr *iph)
887 {
888 struct flow_dissector_key_ip *key_ip;
889
890 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
891 return;
892
893 key_ip = skb_flow_dissector_target(flow_dissector,
894 FLOW_DISSECTOR_KEY_IP,
895 target_container);
896 key_ip->tos = iph->tos;
897 key_ip->ttl = iph->ttl;
898 }
899
900 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)901 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
902 struct flow_dissector *flow_dissector,
903 void *target_container, const void *data,
904 const struct ipv6hdr *iph)
905 {
906 struct flow_dissector_key_ip *key_ip;
907
908 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
909 return;
910
911 key_ip = skb_flow_dissector_target(flow_dissector,
912 FLOW_DISSECTOR_KEY_IP,
913 target_container);
914 key_ip->tos = ipv6_get_dsfield(iph);
915 key_ip->ttl = iph->hop_limit;
916 }
917
918 /* Maximum number of protocol headers that can be parsed in
919 * __skb_flow_dissect
920 */
921 #define MAX_FLOW_DISSECT_HDRS 15
922
skb_flow_dissect_allowed(int * num_hdrs)923 static bool skb_flow_dissect_allowed(int *num_hdrs)
924 {
925 ++*num_hdrs;
926
927 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
928 }
929
__skb_flow_bpf_to_target(const struct bpf_flow_keys * flow_keys,struct flow_dissector * flow_dissector,void * target_container)930 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
931 struct flow_dissector *flow_dissector,
932 void *target_container)
933 {
934 struct flow_dissector_key_ports_range *key_ports_range = NULL;
935 struct flow_dissector_key_ports *key_ports = NULL;
936 struct flow_dissector_key_control *key_control;
937 struct flow_dissector_key_basic *key_basic;
938 struct flow_dissector_key_addrs *key_addrs;
939 struct flow_dissector_key_tags *key_tags;
940
941 key_control = skb_flow_dissector_target(flow_dissector,
942 FLOW_DISSECTOR_KEY_CONTROL,
943 target_container);
944 key_control->thoff = flow_keys->thoff;
945 if (flow_keys->is_frag)
946 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
947 if (flow_keys->is_first_frag)
948 key_control->flags |= FLOW_DIS_FIRST_FRAG;
949 if (flow_keys->is_encap)
950 key_control->flags |= FLOW_DIS_ENCAPSULATION;
951
952 key_basic = skb_flow_dissector_target(flow_dissector,
953 FLOW_DISSECTOR_KEY_BASIC,
954 target_container);
955 key_basic->n_proto = flow_keys->n_proto;
956 key_basic->ip_proto = flow_keys->ip_proto;
957
958 if (flow_keys->addr_proto == ETH_P_IP &&
959 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
960 key_addrs = skb_flow_dissector_target(flow_dissector,
961 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
962 target_container);
963 key_addrs->v4addrs.src = flow_keys->ipv4_src;
964 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
965 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
966 } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
967 dissector_uses_key(flow_dissector,
968 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
969 key_addrs = skb_flow_dissector_target(flow_dissector,
970 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
971 target_container);
972 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
973 sizeof(key_addrs->v6addrs.src));
974 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
975 sizeof(key_addrs->v6addrs.dst));
976 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
977 }
978
979 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) {
980 key_ports = skb_flow_dissector_target(flow_dissector,
981 FLOW_DISSECTOR_KEY_PORTS,
982 target_container);
983 key_ports->src = flow_keys->sport;
984 key_ports->dst = flow_keys->dport;
985 }
986 if (dissector_uses_key(flow_dissector,
987 FLOW_DISSECTOR_KEY_PORTS_RANGE)) {
988 key_ports_range = skb_flow_dissector_target(flow_dissector,
989 FLOW_DISSECTOR_KEY_PORTS_RANGE,
990 target_container);
991 key_ports_range->tp.src = flow_keys->sport;
992 key_ports_range->tp.dst = flow_keys->dport;
993 }
994
995 if (dissector_uses_key(flow_dissector,
996 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
997 key_tags = skb_flow_dissector_target(flow_dissector,
998 FLOW_DISSECTOR_KEY_FLOW_LABEL,
999 target_container);
1000 key_tags->flow_label = ntohl(flow_keys->flow_label);
1001 }
1002 }
1003
bpf_flow_dissect(struct bpf_prog * prog,struct bpf_flow_dissector * ctx,__be16 proto,int nhoff,int hlen,unsigned int flags)1004 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1005 __be16 proto, int nhoff, int hlen, unsigned int flags)
1006 {
1007 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
1008 u32 result;
1009
1010 /* Pass parameters to the BPF program */
1011 memset(flow_keys, 0, sizeof(*flow_keys));
1012 flow_keys->n_proto = proto;
1013 flow_keys->nhoff = nhoff;
1014 flow_keys->thoff = flow_keys->nhoff;
1015
1016 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
1017 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
1018 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
1019 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1020 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
1021 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
1022 flow_keys->flags = flags;
1023
1024 result = bpf_prog_run_pin_on_cpu(prog, ctx);
1025
1026 /* bpf_flow_keys offsets are u16: do not let @hlen be truncated. */
1027 hlen = min_t(int, hlen, U16_MAX);
1028 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1029 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1030 flow_keys->nhoff, hlen);
1031
1032 return result;
1033 }
1034
is_pppoe_ses_hdr_valid(const struct pppoe_hdr * hdr)1035 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1036 {
1037 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1038 }
1039
1040 /**
1041 * __skb_flow_dissect - extract the flow_keys struct and return it
1042 * @net: associated network namespace, derived from @skb if NULL
1043 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1044 * @flow_dissector: list of keys to dissect
1045 * @target_container: target structure to put dissected values into
1046 * @data: raw buffer pointer to the packet, if NULL use skb->data
1047 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1048 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1049 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1050 * @flags: flags that control the dissection process, e.g.
1051 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1052 *
1053 * The function will try to retrieve individual keys into target specified
1054 * by flow_dissector from either the skbuff or a raw buffer specified by the
1055 * rest parameters.
1056 *
1057 * Caller must take care of zeroing target container memory.
1058 */
__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)1059 bool __skb_flow_dissect(const struct net *net,
1060 const struct sk_buff *skb,
1061 struct flow_dissector *flow_dissector,
1062 void *target_container, const void *data,
1063 __be16 proto, int nhoff, int hlen, unsigned int flags)
1064 {
1065 struct flow_dissector_key_control *key_control;
1066 struct flow_dissector_key_basic *key_basic;
1067 struct flow_dissector_key_addrs *key_addrs;
1068 struct flow_dissector_key_tags *key_tags;
1069 struct flow_dissector_key_vlan *key_vlan;
1070 enum flow_dissect_ret fdret;
1071 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1072 bool mpls_el = false;
1073 int mpls_lse = 0;
1074 int num_hdrs = 0;
1075 u8 ip_proto = 0;
1076 u32 thoff;
1077 bool ret;
1078
1079 if (!data) {
1080 data = skb->data;
1081 proto = skb_vlan_tag_present(skb) ?
1082 skb->vlan_proto : skb->protocol;
1083 nhoff = skb_network_offset(skb);
1084 hlen = skb_headlen(skb);
1085 #if IS_ENABLED(CONFIG_NET_DSA)
1086 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1087 proto == htons(ETH_P_XDSA))) {
1088 struct metadata_dst *md_dst = skb_metadata_dst(skb);
1089 const struct dsa_device_ops *ops;
1090 int offset = 0;
1091
1092 ops = skb->dev->dsa_ptr->tag_ops;
1093 /* Only DSA header taggers break flow dissection */
1094 if (ops->needed_headroom &&
1095 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1096 if (ops->flow_dissect)
1097 ops->flow_dissect(skb, &proto, &offset);
1098 else
1099 dsa_tag_generic_flow_dissect(skb,
1100 &proto,
1101 &offset);
1102 hlen -= offset;
1103 nhoff += offset;
1104 }
1105 }
1106 #endif
1107 }
1108
1109 /* It is ensured by skb_flow_dissector_init() that control key will
1110 * be always present.
1111 */
1112 key_control = skb_flow_dissector_target(flow_dissector,
1113 FLOW_DISSECTOR_KEY_CONTROL,
1114 target_container);
1115
1116 /* It is ensured by skb_flow_dissector_init() that basic key will
1117 * be always present.
1118 */
1119 key_basic = skb_flow_dissector_target(flow_dissector,
1120 FLOW_DISSECTOR_KEY_BASIC,
1121 target_container);
1122
1123 rcu_read_lock();
1124
1125 if (skb) {
1126 if (!net) {
1127 if (skb->dev)
1128 net = dev_net_rcu(skb->dev);
1129 else if (skb->sk)
1130 net = sock_net(skb->sk);
1131 }
1132 }
1133
1134 DEBUG_NET_WARN_ON_ONCE(!net);
1135 if (net) {
1136 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1137 struct bpf_prog_array *run_array;
1138
1139 run_array = rcu_dereference(init_net.bpf.run_array[type]);
1140 if (!run_array)
1141 run_array = rcu_dereference(net->bpf.run_array[type]);
1142
1143 if (run_array) {
1144 struct bpf_flow_keys flow_keys;
1145 struct bpf_flow_dissector ctx = {
1146 .flow_keys = &flow_keys,
1147 .data = data,
1148 .data_end = data + hlen,
1149 };
1150 __be16 n_proto = proto;
1151 struct bpf_prog *prog;
1152 u32 result;
1153
1154 if (skb) {
1155 ctx.skb = skb;
1156 /* we can't use 'proto' in the skb case
1157 * because it might be set to skb->vlan_proto
1158 * which has been pulled from the data
1159 */
1160 n_proto = skb->protocol;
1161 }
1162
1163 prog = READ_ONCE(run_array->items[0].prog);
1164 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1165 hlen, flags);
1166 if (result != BPF_FLOW_DISSECTOR_CONTINUE) {
1167 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1168 target_container);
1169 rcu_read_unlock();
1170 return result == BPF_OK;
1171 }
1172 }
1173 }
1174
1175 rcu_read_unlock();
1176
1177 if (dissector_uses_key(flow_dissector,
1178 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1179 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1180
1181 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1182 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1183 target_container);
1184 /* TC filter blocks can be shared across devices with
1185 * different link types, so we cannot validate this
1186 * when the filter is installed -- check at dissect time.
1187 */
1188 if (skb && skb->dev &&
1189 skb->dev->type == ARPHRD_ETHER &&
1190 skb_mac_header_was_set(skb))
1191 memcpy(key_eth_addrs, eth_hdr(skb), sizeof(*key_eth_addrs));
1192 else
1193 memset(key_eth_addrs, 0, sizeof(*key_eth_addrs));
1194 }
1195
1196 if (dissector_uses_key(flow_dissector,
1197 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1198 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1199
1200 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1201 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1202 target_container);
1203 key_num_of_vlans->num_of_vlans = 0;
1204 }
1205
1206 proto_again:
1207 fdret = FLOW_DISSECT_RET_CONTINUE;
1208
1209 switch (proto) {
1210 case htons(ETH_P_IP): {
1211 const struct iphdr *iph;
1212 struct iphdr _iph;
1213
1214 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1215 if (!iph || iph->ihl < 5) {
1216 fdret = FLOW_DISSECT_RET_OUT_BAD;
1217 break;
1218 }
1219
1220 nhoff += iph->ihl * 4;
1221
1222 ip_proto = iph->protocol;
1223
1224 if (dissector_uses_key(flow_dissector,
1225 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1226 key_addrs = skb_flow_dissector_target(flow_dissector,
1227 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1228 target_container);
1229
1230 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1231 sizeof(key_addrs->v4addrs.src));
1232 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1233 sizeof(key_addrs->v4addrs.dst));
1234 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1235 }
1236
1237 __skb_flow_dissect_ipv4(skb, flow_dissector,
1238 target_container, data, iph);
1239
1240 if (ip_is_fragment(iph)) {
1241 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1242
1243 if (iph->frag_off & htons(IP_OFFSET)) {
1244 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1245 break;
1246 } else {
1247 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1248 if (!(flags &
1249 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1250 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1251 break;
1252 }
1253 }
1254 }
1255
1256 break;
1257 }
1258 case htons(ETH_P_IPV6): {
1259 const struct ipv6hdr *iph;
1260 struct ipv6hdr _iph;
1261
1262 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1263 if (!iph) {
1264 fdret = FLOW_DISSECT_RET_OUT_BAD;
1265 break;
1266 }
1267
1268 ip_proto = iph->nexthdr;
1269 nhoff += sizeof(struct ipv6hdr);
1270
1271 if (dissector_uses_key(flow_dissector,
1272 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1273 key_addrs = skb_flow_dissector_target(flow_dissector,
1274 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1275 target_container);
1276
1277 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1278 sizeof(key_addrs->v6addrs.src));
1279 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1280 sizeof(key_addrs->v6addrs.dst));
1281 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1282 }
1283
1284 if ((dissector_uses_key(flow_dissector,
1285 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1286 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1287 ip6_flowlabel(iph)) {
1288 __be32 flow_label = ip6_flowlabel(iph);
1289
1290 if (dissector_uses_key(flow_dissector,
1291 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1292 key_tags = skb_flow_dissector_target(flow_dissector,
1293 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1294 target_container);
1295 key_tags->flow_label = ntohl(flow_label);
1296 }
1297 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1298 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1299 break;
1300 }
1301 }
1302
1303 __skb_flow_dissect_ipv6(skb, flow_dissector,
1304 target_container, data, iph);
1305
1306 break;
1307 }
1308 case htons(ETH_P_8021AD):
1309 case htons(ETH_P_8021Q): {
1310 const struct vlan_hdr *vlan = NULL;
1311 struct vlan_hdr _vlan;
1312 __be16 saved_vlan_tpid = proto;
1313
1314 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1315 skb && skb_vlan_tag_present(skb)) {
1316 proto = skb->protocol;
1317 } else {
1318 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1319 data, hlen, &_vlan);
1320 if (!vlan) {
1321 fdret = FLOW_DISSECT_RET_OUT_BAD;
1322 break;
1323 }
1324
1325 proto = vlan->h_vlan_encapsulated_proto;
1326 nhoff += sizeof(*vlan);
1327 }
1328
1329 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1330 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1331 struct flow_dissector_key_num_of_vlans *key_nvs;
1332
1333 key_nvs = skb_flow_dissector_target(flow_dissector,
1334 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1335 target_container);
1336 key_nvs->num_of_vlans++;
1337 }
1338
1339 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1340 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1341 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1342 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1343 } else {
1344 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1345 break;
1346 }
1347
1348 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1349 key_vlan = skb_flow_dissector_target(flow_dissector,
1350 dissector_vlan,
1351 target_container);
1352
1353 if (!vlan) {
1354 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1355 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1356 } else {
1357 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1358 VLAN_VID_MASK;
1359 key_vlan->vlan_priority =
1360 (ntohs(vlan->h_vlan_TCI) &
1361 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1362 }
1363 key_vlan->vlan_tpid = saved_vlan_tpid;
1364 key_vlan->vlan_eth_type = proto;
1365 }
1366
1367 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1368 break;
1369 }
1370 case htons(ETH_P_PPP_SES): {
1371 struct {
1372 struct pppoe_hdr hdr;
1373 __be16 proto;
1374 } *hdr, _hdr;
1375 u16 ppp_proto;
1376
1377 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1378 if (!hdr) {
1379 fdret = FLOW_DISSECT_RET_OUT_BAD;
1380 break;
1381 }
1382
1383 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1384 fdret = FLOW_DISSECT_RET_OUT_BAD;
1385 break;
1386 }
1387
1388 /* PFC (compressed 1-byte protocol) frames are not processed.
1389 * A compressed protocol field has the least significant bit of
1390 * the most significant octet set, which will fail the following
1391 * ppp_proto_is_valid(), returning FLOW_DISSECT_RET_OUT_BAD.
1392 */
1393 ppp_proto = ntohs(hdr->proto);
1394 nhoff += PPPOE_SES_HLEN;
1395
1396 if (ppp_proto == PPP_IP) {
1397 proto = htons(ETH_P_IP);
1398 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1399 } else if (ppp_proto == PPP_IPV6) {
1400 proto = htons(ETH_P_IPV6);
1401 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1402 } else if (ppp_proto == PPP_MPLS_UC) {
1403 proto = htons(ETH_P_MPLS_UC);
1404 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1405 } else if (ppp_proto == PPP_MPLS_MC) {
1406 proto = htons(ETH_P_MPLS_MC);
1407 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1408 } else if (ppp_proto_is_valid(ppp_proto)) {
1409 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1410 } else {
1411 fdret = FLOW_DISSECT_RET_OUT_BAD;
1412 break;
1413 }
1414
1415 if (dissector_uses_key(flow_dissector,
1416 FLOW_DISSECTOR_KEY_PPPOE)) {
1417 struct flow_dissector_key_pppoe *key_pppoe;
1418
1419 key_pppoe = skb_flow_dissector_target(flow_dissector,
1420 FLOW_DISSECTOR_KEY_PPPOE,
1421 target_container);
1422 key_pppoe->session_id = hdr->hdr.sid;
1423 key_pppoe->ppp_proto = htons(ppp_proto);
1424 key_pppoe->type = htons(ETH_P_PPP_SES);
1425 }
1426 break;
1427 }
1428 case htons(ETH_P_TIPC): {
1429 struct tipc_basic_hdr *hdr, _hdr;
1430
1431 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1432 data, hlen, &_hdr);
1433 if (!hdr) {
1434 fdret = FLOW_DISSECT_RET_OUT_BAD;
1435 break;
1436 }
1437
1438 if (dissector_uses_key(flow_dissector,
1439 FLOW_DISSECTOR_KEY_TIPC)) {
1440 key_addrs = skb_flow_dissector_target(flow_dissector,
1441 FLOW_DISSECTOR_KEY_TIPC,
1442 target_container);
1443 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1444 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1445 }
1446 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1447 break;
1448 }
1449
1450 case htons(ETH_P_MPLS_UC):
1451 case htons(ETH_P_MPLS_MC):
1452 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1453 target_container, data,
1454 nhoff, hlen, mpls_lse,
1455 &mpls_el);
1456 nhoff += sizeof(struct mpls_label);
1457 mpls_lse++;
1458 break;
1459 case htons(ETH_P_FCOE):
1460 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1461 fdret = FLOW_DISSECT_RET_OUT_BAD;
1462 break;
1463 }
1464
1465 nhoff += FCOE_HEADER_LEN;
1466 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1467 break;
1468
1469 case htons(ETH_P_ARP):
1470 case htons(ETH_P_RARP):
1471 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1472 target_container, data,
1473 nhoff, hlen);
1474 break;
1475
1476 case htons(ETH_P_BATMAN):
1477 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1478 &proto, &nhoff, hlen, flags);
1479 break;
1480
1481 case htons(ETH_P_1588): {
1482 struct ptp_header *hdr, _hdr;
1483
1484 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1485 hlen, &_hdr);
1486 if (!hdr) {
1487 fdret = FLOW_DISSECT_RET_OUT_BAD;
1488 break;
1489 }
1490
1491 nhoff += sizeof(struct ptp_header);
1492 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1493 break;
1494 }
1495
1496 case htons(ETH_P_PRP):
1497 case htons(ETH_P_HSR): {
1498 struct hsr_tag *hdr, _hdr;
1499
1500 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1501 &_hdr);
1502 if (!hdr) {
1503 fdret = FLOW_DISSECT_RET_OUT_BAD;
1504 break;
1505 }
1506
1507 proto = hdr->encap_proto;
1508 nhoff += HSR_HLEN;
1509 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1510 break;
1511 }
1512
1513 case htons(ETH_P_CFM):
1514 fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1515 target_container, data,
1516 nhoff, hlen);
1517 break;
1518
1519 default:
1520 fdret = FLOW_DISSECT_RET_OUT_BAD;
1521 break;
1522 }
1523
1524 /* Process result of proto processing */
1525 switch (fdret) {
1526 case FLOW_DISSECT_RET_OUT_GOOD:
1527 goto out_good;
1528 case FLOW_DISSECT_RET_PROTO_AGAIN:
1529 if (skb_flow_dissect_allowed(&num_hdrs))
1530 goto proto_again;
1531 goto out_good;
1532 case FLOW_DISSECT_RET_CONTINUE:
1533 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1534 break;
1535 case FLOW_DISSECT_RET_OUT_BAD:
1536 default:
1537 goto out_bad;
1538 }
1539
1540 ip_proto_again:
1541 fdret = FLOW_DISSECT_RET_CONTINUE;
1542
1543 switch (ip_proto) {
1544 case IPPROTO_GRE:
1545 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1546 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1547 break;
1548 }
1549
1550 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1551 target_container, data,
1552 &proto, &nhoff, &hlen, flags);
1553 break;
1554
1555 case NEXTHDR_HOP:
1556 case NEXTHDR_ROUTING:
1557 case NEXTHDR_DEST: {
1558 u8 _opthdr[2], *opthdr;
1559
1560 if (proto != htons(ETH_P_IPV6))
1561 break;
1562
1563 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1564 data, hlen, &_opthdr);
1565 if (!opthdr) {
1566 fdret = FLOW_DISSECT_RET_OUT_BAD;
1567 break;
1568 }
1569
1570 ip_proto = opthdr[0];
1571 nhoff += (opthdr[1] + 1) << 3;
1572
1573 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1574 break;
1575 }
1576 case NEXTHDR_FRAGMENT: {
1577 struct frag_hdr _fh, *fh;
1578
1579 if (proto != htons(ETH_P_IPV6))
1580 break;
1581
1582 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1583 data, hlen, &_fh);
1584
1585 if (!fh) {
1586 fdret = FLOW_DISSECT_RET_OUT_BAD;
1587 break;
1588 }
1589
1590 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1591
1592 nhoff += sizeof(_fh);
1593 ip_proto = fh->nexthdr;
1594
1595 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1596 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1597 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1598 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1599 break;
1600 }
1601 }
1602
1603 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1604 break;
1605 }
1606 case IPPROTO_IPIP:
1607 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1608 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1609 break;
1610 }
1611
1612 proto = htons(ETH_P_IP);
1613
1614 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1615 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1616 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1617 break;
1618 }
1619
1620 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1621 break;
1622
1623 case IPPROTO_IPV6:
1624 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1625 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1626 break;
1627 }
1628
1629 proto = htons(ETH_P_IPV6);
1630
1631 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1632 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1633 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1634 break;
1635 }
1636
1637 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1638 break;
1639
1640
1641 case IPPROTO_MPLS:
1642 proto = htons(ETH_P_MPLS_UC);
1643 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1644 break;
1645
1646 case IPPROTO_TCP:
1647 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1648 data, nhoff, hlen);
1649 break;
1650
1651 case IPPROTO_ICMP:
1652 case IPPROTO_ICMPV6:
1653 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1654 data, nhoff, hlen);
1655 break;
1656 case IPPROTO_L2TP:
1657 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1658 data, nhoff, hlen);
1659 break;
1660 case IPPROTO_ESP:
1661 __skb_flow_dissect_esp(skb, flow_dissector, target_container,
1662 data, nhoff, hlen);
1663 break;
1664 case IPPROTO_AH:
1665 __skb_flow_dissect_ah(skb, flow_dissector, target_container,
1666 data, nhoff, hlen);
1667 break;
1668 default:
1669 break;
1670 }
1671
1672 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1673 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1674 data, nhoff, ip_proto, hlen);
1675
1676 /* Process result of IP proto processing */
1677 switch (fdret) {
1678 case FLOW_DISSECT_RET_PROTO_AGAIN:
1679 if (skb_flow_dissect_allowed(&num_hdrs))
1680 goto proto_again;
1681 break;
1682 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1683 if (skb_flow_dissect_allowed(&num_hdrs))
1684 goto ip_proto_again;
1685 break;
1686 case FLOW_DISSECT_RET_OUT_GOOD:
1687 case FLOW_DISSECT_RET_CONTINUE:
1688 break;
1689 case FLOW_DISSECT_RET_OUT_BAD:
1690 default:
1691 goto out_bad;
1692 }
1693
1694 out_good:
1695 ret = true;
1696
1697 out:
1698 thoff = min_t(u32, nhoff, skb ? skb->len : hlen);
1699 if (unlikely(thoff > U16_MAX)) {
1700 /* Cannot be represented in key_control->thoff. */
1701 thoff = U16_MAX;
1702 ret = false;
1703 }
1704 key_control->thoff = thoff;
1705 key_basic->n_proto = proto;
1706 key_basic->ip_proto = ip_proto;
1707
1708 return ret;
1709
1710 out_bad:
1711 ret = false;
1712 goto out;
1713 }
1714 EXPORT_SYMBOL(__skb_flow_dissect);
1715
1716 static siphash_aligned_key_t hashrnd;
__flow_hash_secret_init(void)1717 static __always_inline void __flow_hash_secret_init(void)
1718 {
1719 net_get_random_once(&hashrnd, sizeof(hashrnd));
1720 }
1721
flow_keys_hash_start(const struct flow_keys * flow)1722 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1723 {
1724 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1725 return &flow->FLOW_KEYS_HASH_START_FIELD;
1726 }
1727
flow_keys_hash_length(const struct flow_keys * flow)1728 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1729 {
1730 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1731
1732 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1733
1734 switch (flow->control.addr_type) {
1735 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1736 diff -= sizeof(flow->addrs.v4addrs);
1737 break;
1738 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1739 diff -= sizeof(flow->addrs.v6addrs);
1740 break;
1741 case FLOW_DISSECTOR_KEY_TIPC:
1742 diff -= sizeof(flow->addrs.tipckey);
1743 break;
1744 }
1745 return sizeof(*flow) - diff;
1746 }
1747
flow_get_u32_src(const struct flow_keys * flow)1748 __be32 flow_get_u32_src(const struct flow_keys *flow)
1749 {
1750 switch (flow->control.addr_type) {
1751 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1752 return flow->addrs.v4addrs.src;
1753 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1754 return (__force __be32)ipv6_addr_hash(
1755 &flow->addrs.v6addrs.src);
1756 case FLOW_DISSECTOR_KEY_TIPC:
1757 return flow->addrs.tipckey.key;
1758 default:
1759 return 0;
1760 }
1761 }
1762 EXPORT_SYMBOL(flow_get_u32_src);
1763
flow_get_u32_dst(const struct flow_keys * flow)1764 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1765 {
1766 switch (flow->control.addr_type) {
1767 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1768 return flow->addrs.v4addrs.dst;
1769 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1770 return (__force __be32)ipv6_addr_hash(
1771 &flow->addrs.v6addrs.dst);
1772 default:
1773 return 0;
1774 }
1775 }
1776 EXPORT_SYMBOL(flow_get_u32_dst);
1777
1778 /* Sort the source and destination IP and the ports,
1779 * to have consistent hash within the two directions
1780 */
__flow_hash_consistentify(struct flow_keys * keys)1781 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1782 {
1783 int addr_diff, i;
1784
1785 switch (keys->control.addr_type) {
1786 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1787 if ((__force u32)keys->addrs.v4addrs.dst <
1788 (__force u32)keys->addrs.v4addrs.src)
1789 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1790
1791 if ((__force u16)keys->ports.dst <
1792 (__force u16)keys->ports.src) {
1793 swap(keys->ports.src, keys->ports.dst);
1794 }
1795 break;
1796 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1797 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1798 &keys->addrs.v6addrs.src,
1799 sizeof(keys->addrs.v6addrs.dst));
1800 if (addr_diff < 0) {
1801 for (i = 0; i < 4; i++)
1802 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1803 keys->addrs.v6addrs.dst.s6_addr32[i]);
1804 }
1805 if ((__force u16)keys->ports.dst <
1806 (__force u16)keys->ports.src) {
1807 swap(keys->ports.src, keys->ports.dst);
1808 }
1809 break;
1810 }
1811 }
1812
__flow_hash_from_keys(struct flow_keys * keys,const siphash_key_t * keyval)1813 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1814 const siphash_key_t *keyval)
1815 {
1816 u32 hash;
1817
1818 __flow_hash_consistentify(keys);
1819
1820 hash = siphash(flow_keys_hash_start(keys),
1821 flow_keys_hash_length(keys), keyval);
1822 if (!hash)
1823 hash = 1;
1824
1825 return hash;
1826 }
1827
flow_hash_from_keys(struct flow_keys * keys)1828 u32 flow_hash_from_keys(struct flow_keys *keys)
1829 {
1830 __flow_hash_secret_init();
1831 return __flow_hash_from_keys(keys, &hashrnd);
1832 }
1833 EXPORT_SYMBOL(flow_hash_from_keys);
1834
flow_hash_from_keys_seed(struct flow_keys * keys,const siphash_key_t * keyval)1835 u32 flow_hash_from_keys_seed(struct flow_keys *keys,
1836 const siphash_key_t *keyval)
1837 {
1838 return __flow_hash_from_keys(keys, keyval);
1839 }
1840 EXPORT_SYMBOL(flow_hash_from_keys_seed);
1841
___skb_get_hash(const struct sk_buff * skb,struct flow_keys * keys,const siphash_key_t * keyval)1842 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1843 struct flow_keys *keys,
1844 const siphash_key_t *keyval)
1845 {
1846 skb_flow_dissect_flow_keys(skb, keys,
1847 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1848
1849 return __flow_hash_from_keys(keys, keyval);
1850 }
1851
1852 struct _flow_keys_digest_data {
1853 __be16 n_proto;
1854 u8 ip_proto;
1855 u8 padding;
1856 __be32 ports;
1857 __be32 src;
1858 __be32 dst;
1859 };
1860
make_flow_keys_digest(struct flow_keys_digest * digest,const struct flow_keys * flow)1861 void make_flow_keys_digest(struct flow_keys_digest *digest,
1862 const struct flow_keys *flow)
1863 {
1864 struct _flow_keys_digest_data *data =
1865 (struct _flow_keys_digest_data *)digest;
1866
1867 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1868
1869 memset(digest, 0, sizeof(*digest));
1870
1871 data->n_proto = flow->basic.n_proto;
1872 data->ip_proto = flow->basic.ip_proto;
1873 data->ports = flow->ports.ports;
1874 data->src = flow->addrs.v4addrs.src;
1875 data->dst = flow->addrs.v4addrs.dst;
1876 }
1877 EXPORT_SYMBOL(make_flow_keys_digest);
1878
1879 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1880
__skb_get_hash_symmetric_net(const struct net * net,const struct sk_buff * skb)1881 u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb)
1882 {
1883 struct flow_keys keys;
1884
1885 __flow_hash_secret_init();
1886
1887 memset(&keys, 0, sizeof(keys));
1888 __skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric,
1889 &keys, NULL, 0, 0, 0, 0);
1890
1891 return __flow_hash_from_keys(&keys, &hashrnd);
1892 }
1893 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net);
1894
1895 /**
1896 * __skb_get_hash_net: calculate a flow hash
1897 * @net: associated network namespace, derived from @skb if NULL
1898 * @skb: sk_buff to calculate flow hash from
1899 *
1900 * This function calculates a flow hash based on src/dst addresses
1901 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1902 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1903 * if hash is a canonical 4-tuple hash over transport ports.
1904 */
__skb_get_hash_net(const struct net * net,struct sk_buff * skb)1905 void __skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1906 {
1907 struct flow_keys keys;
1908 u32 hash;
1909
1910 memset(&keys, 0, sizeof(keys));
1911
1912 __skb_flow_dissect(net, skb, &flow_keys_dissector,
1913 &keys, NULL, 0, 0, 0,
1914 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1915
1916 __flow_hash_secret_init();
1917
1918 hash = __flow_hash_from_keys(&keys, &hashrnd);
1919
1920 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1921 }
1922 EXPORT_SYMBOL(__skb_get_hash_net);
1923
skb_get_hash_perturb(const struct sk_buff * skb,const siphash_key_t * perturb)1924 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1925 const siphash_key_t *perturb)
1926 {
1927 struct flow_keys keys;
1928
1929 return ___skb_get_hash(skb, &keys, perturb);
1930 }
1931 EXPORT_SYMBOL(skb_get_hash_perturb);
1932
__skb_get_poff(const struct sk_buff * skb,const void * data,const struct flow_keys_basic * keys,int hlen)1933 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1934 const struct flow_keys_basic *keys, int hlen)
1935 {
1936 u32 poff = keys->control.thoff;
1937
1938 /* skip L4 headers for fragments after the first */
1939 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1940 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1941 return poff;
1942
1943 switch (keys->basic.ip_proto) {
1944 case IPPROTO_TCP: {
1945 /* access doff as u8 to avoid unaligned access */
1946 const u8 *doff;
1947 u8 _doff;
1948
1949 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1950 data, hlen, &_doff);
1951 if (!doff)
1952 return poff;
1953
1954 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1955 break;
1956 }
1957 case IPPROTO_UDP:
1958 case IPPROTO_UDPLITE:
1959 poff += sizeof(struct udphdr);
1960 break;
1961 /* For the rest, we do not really care about header
1962 * extensions at this point for now.
1963 */
1964 case IPPROTO_ICMP:
1965 poff += sizeof(struct icmphdr);
1966 break;
1967 case IPPROTO_ICMPV6:
1968 poff += sizeof(struct icmp6hdr);
1969 break;
1970 case IPPROTO_IGMP:
1971 poff += sizeof(struct igmphdr);
1972 break;
1973 case IPPROTO_DCCP:
1974 poff += sizeof(struct dccp_hdr);
1975 break;
1976 case IPPROTO_SCTP:
1977 poff += sizeof(struct sctphdr);
1978 break;
1979 }
1980
1981 return poff;
1982 }
1983
1984 /**
1985 * skb_get_poff - get the offset to the payload
1986 * @skb: sk_buff to get the payload offset from
1987 *
1988 * The function will get the offset to the payload as far as it could
1989 * be dissected. The main user is currently BPF, so that we can dynamically
1990 * truncate packets without needing to push actual payload to the user
1991 * space and can analyze headers only, instead.
1992 */
skb_get_poff(const struct sk_buff * skb)1993 u32 skb_get_poff(const struct sk_buff *skb)
1994 {
1995 struct flow_keys_basic keys;
1996
1997 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1998 NULL, 0, 0, 0, 0))
1999 return 0;
2000
2001 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
2002 }
2003
__get_hash_from_flowi6(const struct flowi6 * fl6,struct flow_keys * keys)2004 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
2005 {
2006 memset(keys, 0, sizeof(*keys));
2007
2008 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
2009 sizeof(keys->addrs.v6addrs.src));
2010 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
2011 sizeof(keys->addrs.v6addrs.dst));
2012 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2013 keys->ports.src = fl6->fl6_sport;
2014 keys->ports.dst = fl6->fl6_dport;
2015 keys->keyid.keyid = fl6->fl6_gre_key;
2016 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2017 keys->basic.ip_proto = fl6->flowi6_proto;
2018
2019 return flow_hash_from_keys(keys);
2020 }
2021 EXPORT_SYMBOL(__get_hash_from_flowi6);
2022
2023 static const struct flow_dissector_key flow_keys_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 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2034 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2035 },
2036 {
2037 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2038 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2039 },
2040 {
2041 .key_id = FLOW_DISSECTOR_KEY_TIPC,
2042 .offset = offsetof(struct flow_keys, addrs.tipckey),
2043 },
2044 {
2045 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2046 .offset = offsetof(struct flow_keys, ports),
2047 },
2048 {
2049 .key_id = FLOW_DISSECTOR_KEY_VLAN,
2050 .offset = offsetof(struct flow_keys, vlan),
2051 },
2052 {
2053 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2054 .offset = offsetof(struct flow_keys, tags),
2055 },
2056 {
2057 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2058 .offset = offsetof(struct flow_keys, keyid),
2059 },
2060 };
2061
2062 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2063 {
2064 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2065 .offset = offsetof(struct flow_keys, control),
2066 },
2067 {
2068 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2069 .offset = offsetof(struct flow_keys, basic),
2070 },
2071 {
2072 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2073 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2074 },
2075 {
2076 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2077 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2078 },
2079 {
2080 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2081 .offset = offsetof(struct flow_keys, ports),
2082 },
2083 };
2084
2085 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2086 {
2087 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2088 .offset = offsetof(struct flow_keys, control),
2089 },
2090 {
2091 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2092 .offset = offsetof(struct flow_keys, basic),
2093 },
2094 };
2095
2096 struct flow_dissector flow_keys_dissector __read_mostly;
2097 EXPORT_SYMBOL(flow_keys_dissector);
2098
2099 struct flow_dissector flow_keys_basic_dissector __read_mostly;
2100 EXPORT_SYMBOL(flow_keys_basic_dissector);
2101
init_default_flow_dissectors(void)2102 static int __init init_default_flow_dissectors(void)
2103 {
2104 skb_flow_dissector_init(&flow_keys_dissector,
2105 flow_keys_dissector_keys,
2106 ARRAY_SIZE(flow_keys_dissector_keys));
2107 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2108 flow_keys_dissector_symmetric_keys,
2109 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2110 skb_flow_dissector_init(&flow_keys_basic_dissector,
2111 flow_keys_basic_dissector_keys,
2112 ARRAY_SIZE(flow_keys_basic_dissector_keys));
2113 return 0;
2114 }
2115 core_initcall(init_default_flow_dissectors);
2116