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 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
459 u32 val;
460
461 enc_opt = skb_flow_dissector_target(flow_dissector,
462 FLOW_DISSECTOR_KEY_ENC_OPTS,
463 target_container);
464
465 if (!info->options_len)
466 return;
467
468 enc_opt->len = info->options_len;
469 ip_tunnel_info_opts_get(enc_opt->data, info);
470
471 ip_tunnel_set_options_present(flags);
472 ip_tunnel_flags_and(flags, info->key.tun_flags, flags);
473
474 val = find_next_bit(flags, __IP_TUNNEL_FLAG_NUM,
475 IP_TUNNEL_GENEVE_OPT_BIT);
476 enc_opt->dst_opt_type = val < __IP_TUNNEL_FLAG_NUM ? val : 0;
477 }
478 }
479 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
480
skb_flow_dissect_hash(const struct sk_buff * skb,struct flow_dissector * flow_dissector,void * target_container)481 void skb_flow_dissect_hash(const struct sk_buff *skb,
482 struct flow_dissector *flow_dissector,
483 void *target_container)
484 {
485 struct flow_dissector_key_hash *key;
486
487 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
488 return;
489
490 key = skb_flow_dissector_target(flow_dissector,
491 FLOW_DISSECTOR_KEY_HASH,
492 target_container);
493
494 key->hash = skb_get_hash_raw(skb);
495 }
496 EXPORT_SYMBOL(skb_flow_dissect_hash);
497
498 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)499 __skb_flow_dissect_mpls(const struct sk_buff *skb,
500 struct flow_dissector *flow_dissector,
501 void *target_container, const void *data, int nhoff,
502 int hlen, int lse_index, bool *entropy_label)
503 {
504 struct mpls_label *hdr, _hdr;
505 u32 entry, label, bos;
506
507 if (!dissector_uses_key(flow_dissector,
508 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
509 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
510 return FLOW_DISSECT_RET_OUT_GOOD;
511
512 if (lse_index >= FLOW_DIS_MPLS_MAX)
513 return FLOW_DISSECT_RET_OUT_GOOD;
514
515 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
516 hlen, &_hdr);
517 if (!hdr)
518 return FLOW_DISSECT_RET_OUT_BAD;
519
520 entry = ntohl(hdr->entry);
521 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
522 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
523
524 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
525 struct flow_dissector_key_mpls *key_mpls;
526 struct flow_dissector_mpls_lse *lse;
527
528 key_mpls = skb_flow_dissector_target(flow_dissector,
529 FLOW_DISSECTOR_KEY_MPLS,
530 target_container);
531 lse = &key_mpls->ls[lse_index];
532
533 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
534 lse->mpls_bos = bos;
535 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
536 lse->mpls_label = label;
537 dissector_set_mpls_lse(key_mpls, lse_index);
538 }
539
540 if (*entropy_label &&
541 dissector_uses_key(flow_dissector,
542 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
543 struct flow_dissector_key_keyid *key_keyid;
544
545 key_keyid = skb_flow_dissector_target(flow_dissector,
546 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
547 target_container);
548 key_keyid->keyid = cpu_to_be32(label);
549 }
550
551 *entropy_label = label == MPLS_LABEL_ENTROPY;
552
553 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
554 }
555
556 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)557 __skb_flow_dissect_arp(const struct sk_buff *skb,
558 struct flow_dissector *flow_dissector,
559 void *target_container, const void *data,
560 int nhoff, int hlen)
561 {
562 struct flow_dissector_key_arp *key_arp;
563 struct {
564 unsigned char ar_sha[ETH_ALEN];
565 unsigned char ar_sip[4];
566 unsigned char ar_tha[ETH_ALEN];
567 unsigned char ar_tip[4];
568 } *arp_eth, _arp_eth;
569 const struct arphdr *arp;
570 struct arphdr _arp;
571
572 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
573 return FLOW_DISSECT_RET_OUT_GOOD;
574
575 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
576 hlen, &_arp);
577 if (!arp)
578 return FLOW_DISSECT_RET_OUT_BAD;
579
580 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
581 arp->ar_pro != htons(ETH_P_IP) ||
582 arp->ar_hln != ETH_ALEN ||
583 arp->ar_pln != 4 ||
584 (arp->ar_op != htons(ARPOP_REPLY) &&
585 arp->ar_op != htons(ARPOP_REQUEST)))
586 return FLOW_DISSECT_RET_OUT_BAD;
587
588 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
589 sizeof(_arp_eth), data,
590 hlen, &_arp_eth);
591 if (!arp_eth)
592 return FLOW_DISSECT_RET_OUT_BAD;
593
594 key_arp = skb_flow_dissector_target(flow_dissector,
595 FLOW_DISSECTOR_KEY_ARP,
596 target_container);
597
598 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
599 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
600
601 /* Only store the lower byte of the opcode;
602 * this covers ARPOP_REPLY and ARPOP_REQUEST.
603 */
604 key_arp->op = ntohs(arp->ar_op) & 0xff;
605
606 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
607 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
608
609 return FLOW_DISSECT_RET_OUT_GOOD;
610 }
611
612 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)613 __skb_flow_dissect_cfm(const struct sk_buff *skb,
614 struct flow_dissector *flow_dissector,
615 void *target_container, const void *data,
616 int nhoff, int hlen)
617 {
618 struct flow_dissector_key_cfm *key, *hdr, _hdr;
619
620 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CFM))
621 return FLOW_DISSECT_RET_OUT_GOOD;
622
623 hdr = __skb_header_pointer(skb, nhoff, sizeof(*key), data, hlen, &_hdr);
624 if (!hdr)
625 return FLOW_DISSECT_RET_OUT_BAD;
626
627 key = skb_flow_dissector_target(flow_dissector, FLOW_DISSECTOR_KEY_CFM,
628 target_container);
629
630 key->mdl_ver = hdr->mdl_ver;
631 key->opcode = hdr->opcode;
632
633 return FLOW_DISSECT_RET_OUT_GOOD;
634 }
635
636 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)637 __skb_flow_dissect_gre(const struct sk_buff *skb,
638 struct flow_dissector_key_control *key_control,
639 struct flow_dissector *flow_dissector,
640 void *target_container, const void *data,
641 __be16 *p_proto, int *p_nhoff, int *p_hlen,
642 unsigned int flags)
643 {
644 struct flow_dissector_key_keyid *key_keyid;
645 struct gre_base_hdr *hdr, _hdr;
646 int offset = 0;
647 u16 gre_ver;
648
649 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
650 data, *p_hlen, &_hdr);
651 if (!hdr)
652 return FLOW_DISSECT_RET_OUT_BAD;
653
654 /* Only look inside GRE without routing */
655 if (hdr->flags & GRE_ROUTING)
656 return FLOW_DISSECT_RET_OUT_GOOD;
657
658 /* Only look inside GRE for version 0 and 1 */
659 gre_ver = ntohs(hdr->flags & GRE_VERSION);
660 if (gre_ver > 1)
661 return FLOW_DISSECT_RET_OUT_GOOD;
662
663 *p_proto = hdr->protocol;
664 if (gre_ver) {
665 /* Version1 must be PPTP, and check the flags */
666 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
667 return FLOW_DISSECT_RET_OUT_GOOD;
668 }
669
670 offset += sizeof(struct gre_base_hdr);
671
672 if (hdr->flags & GRE_CSUM)
673 offset += sizeof_field(struct gre_full_hdr, csum) +
674 sizeof_field(struct gre_full_hdr, reserved1);
675
676 if (hdr->flags & GRE_KEY) {
677 const __be32 *keyid;
678 __be32 _keyid;
679
680 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
681 sizeof(_keyid),
682 data, *p_hlen, &_keyid);
683 if (!keyid)
684 return FLOW_DISSECT_RET_OUT_BAD;
685
686 if (dissector_uses_key(flow_dissector,
687 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
688 key_keyid = skb_flow_dissector_target(flow_dissector,
689 FLOW_DISSECTOR_KEY_GRE_KEYID,
690 target_container);
691 if (gre_ver == 0)
692 key_keyid->keyid = *keyid;
693 else
694 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
695 }
696 offset += sizeof_field(struct gre_full_hdr, key);
697 }
698
699 if (hdr->flags & GRE_SEQ)
700 offset += sizeof_field(struct pptp_gre_header, seq);
701
702 if (gre_ver == 0) {
703 if (*p_proto == htons(ETH_P_TEB)) {
704 const struct ethhdr *eth;
705 struct ethhdr _eth;
706
707 eth = __skb_header_pointer(skb, *p_nhoff + offset,
708 sizeof(_eth),
709 data, *p_hlen, &_eth);
710 if (!eth)
711 return FLOW_DISSECT_RET_OUT_BAD;
712 *p_proto = eth->h_proto;
713 offset += sizeof(*eth);
714
715 /* Cap headers that we access via pointers at the
716 * end of the Ethernet header as our maximum alignment
717 * at that point is only 2 bytes.
718 */
719 if (NET_IP_ALIGN)
720 *p_hlen = *p_nhoff + offset;
721 }
722 } else { /* version 1, must be PPTP */
723 u8 _ppp_hdr[PPP_HDRLEN];
724 u8 *ppp_hdr;
725
726 if (hdr->flags & GRE_ACK)
727 offset += sizeof_field(struct pptp_gre_header, ack);
728
729 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
730 sizeof(_ppp_hdr),
731 data, *p_hlen, _ppp_hdr);
732 if (!ppp_hdr)
733 return FLOW_DISSECT_RET_OUT_BAD;
734
735 switch (PPP_PROTOCOL(ppp_hdr)) {
736 case PPP_IP:
737 *p_proto = htons(ETH_P_IP);
738 break;
739 case PPP_IPV6:
740 *p_proto = htons(ETH_P_IPV6);
741 break;
742 default:
743 /* Could probably catch some more like MPLS */
744 break;
745 }
746
747 offset += PPP_HDRLEN;
748 }
749
750 *p_nhoff += offset;
751 key_control->flags |= FLOW_DIS_ENCAPSULATION;
752 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
753 return FLOW_DISSECT_RET_OUT_GOOD;
754
755 return FLOW_DISSECT_RET_PROTO_AGAIN;
756 }
757
758 /**
759 * __skb_flow_dissect_batadv() - dissect batman-adv header
760 * @skb: sk_buff to with the batman-adv header
761 * @key_control: flow dissectors control key
762 * @data: raw buffer pointer to the packet, if NULL use skb->data
763 * @p_proto: pointer used to update the protocol to process next
764 * @p_nhoff: pointer used to update inner network header offset
765 * @hlen: packet header length
766 * @flags: any combination of FLOW_DISSECTOR_F_*
767 *
768 * ETH_P_BATMAN packets are tried to be dissected. Only
769 * &struct batadv_unicast packets are actually processed because they contain an
770 * inner ethernet header and are usually followed by actual network header. This
771 * allows the flow dissector to continue processing the packet.
772 *
773 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
774 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
775 * otherwise FLOW_DISSECT_RET_OUT_BAD
776 */
777 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)778 __skb_flow_dissect_batadv(const struct sk_buff *skb,
779 struct flow_dissector_key_control *key_control,
780 const void *data, __be16 *p_proto, int *p_nhoff,
781 int hlen, unsigned int flags)
782 {
783 struct {
784 struct batadv_unicast_packet batadv_unicast;
785 struct ethhdr eth;
786 } *hdr, _hdr;
787
788 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
789 &_hdr);
790 if (!hdr)
791 return FLOW_DISSECT_RET_OUT_BAD;
792
793 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
794 return FLOW_DISSECT_RET_OUT_BAD;
795
796 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
797 return FLOW_DISSECT_RET_OUT_BAD;
798
799 *p_proto = hdr->eth.h_proto;
800 *p_nhoff += sizeof(*hdr);
801
802 key_control->flags |= FLOW_DIS_ENCAPSULATION;
803 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
804 return FLOW_DISSECT_RET_OUT_GOOD;
805
806 return FLOW_DISSECT_RET_PROTO_AGAIN;
807 }
808
809 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)810 __skb_flow_dissect_tcp(const struct sk_buff *skb,
811 struct flow_dissector *flow_dissector,
812 void *target_container, const void *data,
813 int thoff, int hlen)
814 {
815 struct flow_dissector_key_tcp *key_tcp;
816 struct tcphdr *th, _th;
817
818 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
819 return;
820
821 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
822 if (!th)
823 return;
824
825 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
826 return;
827
828 key_tcp = skb_flow_dissector_target(flow_dissector,
829 FLOW_DISSECTOR_KEY_TCP,
830 target_container);
831 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
832 }
833
834 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)835 __skb_flow_dissect_ports(const struct sk_buff *skb,
836 struct flow_dissector *flow_dissector,
837 void *target_container, const void *data,
838 int nhoff, u8 ip_proto, int hlen)
839 {
840 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
841 struct flow_dissector_key_ports *key_ports;
842
843 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
844 dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
845 else if (dissector_uses_key(flow_dissector,
846 FLOW_DISSECTOR_KEY_PORTS_RANGE))
847 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
848
849 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
850 return;
851
852 key_ports = skb_flow_dissector_target(flow_dissector,
853 dissector_ports,
854 target_container);
855 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
856 data, hlen);
857 }
858
859 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)860 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
861 struct flow_dissector *flow_dissector,
862 void *target_container, const void *data,
863 const struct iphdr *iph)
864 {
865 struct flow_dissector_key_ip *key_ip;
866
867 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
868 return;
869
870 key_ip = skb_flow_dissector_target(flow_dissector,
871 FLOW_DISSECTOR_KEY_IP,
872 target_container);
873 key_ip->tos = iph->tos;
874 key_ip->ttl = iph->ttl;
875 }
876
877 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)878 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
879 struct flow_dissector *flow_dissector,
880 void *target_container, const void *data,
881 const struct ipv6hdr *iph)
882 {
883 struct flow_dissector_key_ip *key_ip;
884
885 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
886 return;
887
888 key_ip = skb_flow_dissector_target(flow_dissector,
889 FLOW_DISSECTOR_KEY_IP,
890 target_container);
891 key_ip->tos = ipv6_get_dsfield(iph);
892 key_ip->ttl = iph->hop_limit;
893 }
894
895 /* Maximum number of protocol headers that can be parsed in
896 * __skb_flow_dissect
897 */
898 #define MAX_FLOW_DISSECT_HDRS 15
899
skb_flow_dissect_allowed(int * num_hdrs)900 static bool skb_flow_dissect_allowed(int *num_hdrs)
901 {
902 ++*num_hdrs;
903
904 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
905 }
906
__skb_flow_bpf_to_target(const struct bpf_flow_keys * flow_keys,struct flow_dissector * flow_dissector,void * target_container)907 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
908 struct flow_dissector *flow_dissector,
909 void *target_container)
910 {
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 else if (dissector_uses_key(flow_dissector,
960 FLOW_DISSECTOR_KEY_PORTS_RANGE))
961 key_ports = skb_flow_dissector_target(flow_dissector,
962 FLOW_DISSECTOR_KEY_PORTS_RANGE,
963 target_container);
964
965 if (key_ports) {
966 key_ports->src = flow_keys->sport;
967 key_ports->dst = flow_keys->dport;
968 }
969
970 if (dissector_uses_key(flow_dissector,
971 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
972 key_tags = skb_flow_dissector_target(flow_dissector,
973 FLOW_DISSECTOR_KEY_FLOW_LABEL,
974 target_container);
975 key_tags->flow_label = ntohl(flow_keys->flow_label);
976 }
977 }
978
bpf_flow_dissect(struct bpf_prog * prog,struct bpf_flow_dissector * ctx,__be16 proto,int nhoff,int hlen,unsigned int flags)979 u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
980 __be16 proto, int nhoff, int hlen, unsigned int flags)
981 {
982 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
983 u32 result;
984
985 /* Pass parameters to the BPF program */
986 memset(flow_keys, 0, sizeof(*flow_keys));
987 flow_keys->n_proto = proto;
988 flow_keys->nhoff = nhoff;
989 flow_keys->thoff = flow_keys->nhoff;
990
991 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
992 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
993 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
994 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
995 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
996 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
997 flow_keys->flags = flags;
998
999 result = bpf_prog_run_pin_on_cpu(prog, ctx);
1000
1001 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1002 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1003 flow_keys->nhoff, hlen);
1004
1005 return result;
1006 }
1007
is_pppoe_ses_hdr_valid(const struct pppoe_hdr * hdr)1008 static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1009 {
1010 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1011 }
1012
1013 /**
1014 * __skb_flow_dissect - extract the flow_keys struct and return it
1015 * @net: associated network namespace, derived from @skb if NULL
1016 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1017 * @flow_dissector: list of keys to dissect
1018 * @target_container: target structure to put dissected values into
1019 * @data: raw buffer pointer to the packet, if NULL use skb->data
1020 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1021 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1022 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1023 * @flags: flags that control the dissection process, e.g.
1024 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1025 *
1026 * The function will try to retrieve individual keys into target specified
1027 * by flow_dissector from either the skbuff or a raw buffer specified by the
1028 * rest parameters.
1029 *
1030 * Caller must take care of zeroing target container memory.
1031 */
__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)1032 bool __skb_flow_dissect(const struct net *net,
1033 const struct sk_buff *skb,
1034 struct flow_dissector *flow_dissector,
1035 void *target_container, const void *data,
1036 __be16 proto, int nhoff, int hlen, unsigned int flags)
1037 {
1038 struct flow_dissector_key_control *key_control;
1039 struct flow_dissector_key_basic *key_basic;
1040 struct flow_dissector_key_addrs *key_addrs;
1041 struct flow_dissector_key_tags *key_tags;
1042 struct flow_dissector_key_vlan *key_vlan;
1043 enum flow_dissect_ret fdret;
1044 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1045 bool mpls_el = false;
1046 int mpls_lse = 0;
1047 int num_hdrs = 0;
1048 u8 ip_proto = 0;
1049 bool ret;
1050
1051 if (!data) {
1052 data = skb->data;
1053 proto = skb_vlan_tag_present(skb) ?
1054 skb->vlan_proto : skb->protocol;
1055 nhoff = skb_network_offset(skb);
1056 hlen = skb_headlen(skb);
1057 #if IS_ENABLED(CONFIG_NET_DSA)
1058 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1059 proto == htons(ETH_P_XDSA))) {
1060 struct metadata_dst *md_dst = skb_metadata_dst(skb);
1061 const struct dsa_device_ops *ops;
1062 int offset = 0;
1063
1064 ops = skb->dev->dsa_ptr->tag_ops;
1065 /* Only DSA header taggers break flow dissection */
1066 if (ops->needed_headroom &&
1067 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1068 if (ops->flow_dissect)
1069 ops->flow_dissect(skb, &proto, &offset);
1070 else
1071 dsa_tag_generic_flow_dissect(skb,
1072 &proto,
1073 &offset);
1074 hlen -= offset;
1075 nhoff += offset;
1076 }
1077 }
1078 #endif
1079 }
1080
1081 /* It is ensured by skb_flow_dissector_init() that control key will
1082 * be always present.
1083 */
1084 key_control = skb_flow_dissector_target(flow_dissector,
1085 FLOW_DISSECTOR_KEY_CONTROL,
1086 target_container);
1087
1088 /* It is ensured by skb_flow_dissector_init() that basic key will
1089 * be always present.
1090 */
1091 key_basic = skb_flow_dissector_target(flow_dissector,
1092 FLOW_DISSECTOR_KEY_BASIC,
1093 target_container);
1094
1095 if (skb) {
1096 if (!net) {
1097 if (skb->dev)
1098 net = dev_net(skb->dev);
1099 else if (skb->sk)
1100 net = sock_net(skb->sk);
1101 }
1102 }
1103
1104 WARN_ON_ONCE(!net);
1105 if (net) {
1106 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1107 struct bpf_prog_array *run_array;
1108
1109 rcu_read_lock();
1110 run_array = rcu_dereference(init_net.bpf.run_array[type]);
1111 if (!run_array)
1112 run_array = rcu_dereference(net->bpf.run_array[type]);
1113
1114 if (run_array) {
1115 struct bpf_flow_keys flow_keys;
1116 struct bpf_flow_dissector ctx = {
1117 .flow_keys = &flow_keys,
1118 .data = data,
1119 .data_end = data + hlen,
1120 };
1121 __be16 n_proto = proto;
1122 struct bpf_prog *prog;
1123 u32 result;
1124
1125 if (skb) {
1126 ctx.skb = skb;
1127 /* we can't use 'proto' in the skb case
1128 * because it might be set to skb->vlan_proto
1129 * which has been pulled from the data
1130 */
1131 n_proto = skb->protocol;
1132 }
1133
1134 prog = READ_ONCE(run_array->items[0].prog);
1135 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1136 hlen, flags);
1137 if (result == BPF_FLOW_DISSECTOR_CONTINUE)
1138 goto dissect_continue;
1139 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1140 target_container);
1141 rcu_read_unlock();
1142 return result == BPF_OK;
1143 }
1144 dissect_continue:
1145 rcu_read_unlock();
1146 }
1147
1148 if (dissector_uses_key(flow_dissector,
1149 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1150 struct ethhdr *eth = eth_hdr(skb);
1151 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1152
1153 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1154 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1155 target_container);
1156 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1157 }
1158
1159 if (dissector_uses_key(flow_dissector,
1160 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1161 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1162
1163 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1164 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1165 target_container);
1166 key_num_of_vlans->num_of_vlans = 0;
1167 }
1168
1169 proto_again:
1170 fdret = FLOW_DISSECT_RET_CONTINUE;
1171
1172 switch (proto) {
1173 case htons(ETH_P_IP): {
1174 const struct iphdr *iph;
1175 struct iphdr _iph;
1176
1177 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1178 if (!iph || iph->ihl < 5) {
1179 fdret = FLOW_DISSECT_RET_OUT_BAD;
1180 break;
1181 }
1182
1183 nhoff += iph->ihl * 4;
1184
1185 ip_proto = iph->protocol;
1186
1187 if (dissector_uses_key(flow_dissector,
1188 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1189 key_addrs = skb_flow_dissector_target(flow_dissector,
1190 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1191 target_container);
1192
1193 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1194 sizeof(key_addrs->v4addrs.src));
1195 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1196 sizeof(key_addrs->v4addrs.dst));
1197 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1198 }
1199
1200 __skb_flow_dissect_ipv4(skb, flow_dissector,
1201 target_container, data, iph);
1202
1203 if (ip_is_fragment(iph)) {
1204 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1205
1206 if (iph->frag_off & htons(IP_OFFSET)) {
1207 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1208 break;
1209 } else {
1210 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1211 if (!(flags &
1212 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1213 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1214 break;
1215 }
1216 }
1217 }
1218
1219 break;
1220 }
1221 case htons(ETH_P_IPV6): {
1222 const struct ipv6hdr *iph;
1223 struct ipv6hdr _iph;
1224
1225 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1226 if (!iph) {
1227 fdret = FLOW_DISSECT_RET_OUT_BAD;
1228 break;
1229 }
1230
1231 ip_proto = iph->nexthdr;
1232 nhoff += sizeof(struct ipv6hdr);
1233
1234 if (dissector_uses_key(flow_dissector,
1235 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1236 key_addrs = skb_flow_dissector_target(flow_dissector,
1237 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1238 target_container);
1239
1240 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1241 sizeof(key_addrs->v6addrs.src));
1242 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1243 sizeof(key_addrs->v6addrs.dst));
1244 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1245 }
1246
1247 if ((dissector_uses_key(flow_dissector,
1248 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1249 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1250 ip6_flowlabel(iph)) {
1251 __be32 flow_label = ip6_flowlabel(iph);
1252
1253 if (dissector_uses_key(flow_dissector,
1254 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1255 key_tags = skb_flow_dissector_target(flow_dissector,
1256 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1257 target_container);
1258 key_tags->flow_label = ntohl(flow_label);
1259 }
1260 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1261 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1262 break;
1263 }
1264 }
1265
1266 __skb_flow_dissect_ipv6(skb, flow_dissector,
1267 target_container, data, iph);
1268
1269 break;
1270 }
1271 case htons(ETH_P_8021AD):
1272 case htons(ETH_P_8021Q): {
1273 const struct vlan_hdr *vlan = NULL;
1274 struct vlan_hdr _vlan;
1275 __be16 saved_vlan_tpid = proto;
1276
1277 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1278 skb && skb_vlan_tag_present(skb)) {
1279 proto = skb->protocol;
1280 } else {
1281 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1282 data, hlen, &_vlan);
1283 if (!vlan) {
1284 fdret = FLOW_DISSECT_RET_OUT_BAD;
1285 break;
1286 }
1287
1288 proto = vlan->h_vlan_encapsulated_proto;
1289 nhoff += sizeof(*vlan);
1290 }
1291
1292 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1293 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1294 struct flow_dissector_key_num_of_vlans *key_nvs;
1295
1296 key_nvs = skb_flow_dissector_target(flow_dissector,
1297 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1298 target_container);
1299 key_nvs->num_of_vlans++;
1300 }
1301
1302 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1303 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1304 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1305 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1306 } else {
1307 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1308 break;
1309 }
1310
1311 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1312 key_vlan = skb_flow_dissector_target(flow_dissector,
1313 dissector_vlan,
1314 target_container);
1315
1316 if (!vlan) {
1317 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1318 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1319 } else {
1320 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1321 VLAN_VID_MASK;
1322 key_vlan->vlan_priority =
1323 (ntohs(vlan->h_vlan_TCI) &
1324 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1325 }
1326 key_vlan->vlan_tpid = saved_vlan_tpid;
1327 key_vlan->vlan_eth_type = proto;
1328 }
1329
1330 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1331 break;
1332 }
1333 case htons(ETH_P_PPP_SES): {
1334 struct {
1335 struct pppoe_hdr hdr;
1336 __be16 proto;
1337 } *hdr, _hdr;
1338 u16 ppp_proto;
1339
1340 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1341 if (!hdr) {
1342 fdret = FLOW_DISSECT_RET_OUT_BAD;
1343 break;
1344 }
1345
1346 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1347 fdret = FLOW_DISSECT_RET_OUT_BAD;
1348 break;
1349 }
1350
1351 /* least significant bit of the most significant octet
1352 * indicates if protocol field was compressed
1353 */
1354 ppp_proto = ntohs(hdr->proto);
1355 if (ppp_proto & 0x0100) {
1356 ppp_proto = ppp_proto >> 8;
1357 nhoff += PPPOE_SES_HLEN - 1;
1358 } else {
1359 nhoff += PPPOE_SES_HLEN;
1360 }
1361
1362 if (ppp_proto == PPP_IP) {
1363 proto = htons(ETH_P_IP);
1364 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1365 } else if (ppp_proto == PPP_IPV6) {
1366 proto = htons(ETH_P_IPV6);
1367 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1368 } else if (ppp_proto == PPP_MPLS_UC) {
1369 proto = htons(ETH_P_MPLS_UC);
1370 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1371 } else if (ppp_proto == PPP_MPLS_MC) {
1372 proto = htons(ETH_P_MPLS_MC);
1373 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1374 } else if (ppp_proto_is_valid(ppp_proto)) {
1375 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1376 } else {
1377 fdret = FLOW_DISSECT_RET_OUT_BAD;
1378 break;
1379 }
1380
1381 if (dissector_uses_key(flow_dissector,
1382 FLOW_DISSECTOR_KEY_PPPOE)) {
1383 struct flow_dissector_key_pppoe *key_pppoe;
1384
1385 key_pppoe = skb_flow_dissector_target(flow_dissector,
1386 FLOW_DISSECTOR_KEY_PPPOE,
1387 target_container);
1388 key_pppoe->session_id = hdr->hdr.sid;
1389 key_pppoe->ppp_proto = htons(ppp_proto);
1390 key_pppoe->type = htons(ETH_P_PPP_SES);
1391 }
1392 break;
1393 }
1394 case htons(ETH_P_TIPC): {
1395 struct tipc_basic_hdr *hdr, _hdr;
1396
1397 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1398 data, hlen, &_hdr);
1399 if (!hdr) {
1400 fdret = FLOW_DISSECT_RET_OUT_BAD;
1401 break;
1402 }
1403
1404 if (dissector_uses_key(flow_dissector,
1405 FLOW_DISSECTOR_KEY_TIPC)) {
1406 key_addrs = skb_flow_dissector_target(flow_dissector,
1407 FLOW_DISSECTOR_KEY_TIPC,
1408 target_container);
1409 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1410 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1411 }
1412 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1413 break;
1414 }
1415
1416 case htons(ETH_P_MPLS_UC):
1417 case htons(ETH_P_MPLS_MC):
1418 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1419 target_container, data,
1420 nhoff, hlen, mpls_lse,
1421 &mpls_el);
1422 nhoff += sizeof(struct mpls_label);
1423 mpls_lse++;
1424 break;
1425 case htons(ETH_P_FCOE):
1426 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1427 fdret = FLOW_DISSECT_RET_OUT_BAD;
1428 break;
1429 }
1430
1431 nhoff += FCOE_HEADER_LEN;
1432 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1433 break;
1434
1435 case htons(ETH_P_ARP):
1436 case htons(ETH_P_RARP):
1437 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1438 target_container, data,
1439 nhoff, hlen);
1440 break;
1441
1442 case htons(ETH_P_BATMAN):
1443 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1444 &proto, &nhoff, hlen, flags);
1445 break;
1446
1447 case htons(ETH_P_1588): {
1448 struct ptp_header *hdr, _hdr;
1449
1450 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1451 hlen, &_hdr);
1452 if (!hdr) {
1453 fdret = FLOW_DISSECT_RET_OUT_BAD;
1454 break;
1455 }
1456
1457 nhoff += sizeof(struct ptp_header);
1458 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1459 break;
1460 }
1461
1462 case htons(ETH_P_PRP):
1463 case htons(ETH_P_HSR): {
1464 struct hsr_tag *hdr, _hdr;
1465
1466 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1467 &_hdr);
1468 if (!hdr) {
1469 fdret = FLOW_DISSECT_RET_OUT_BAD;
1470 break;
1471 }
1472
1473 proto = hdr->encap_proto;
1474 nhoff += HSR_HLEN;
1475 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1476 break;
1477 }
1478
1479 case htons(ETH_P_CFM):
1480 fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1481 target_container, data,
1482 nhoff, hlen);
1483 break;
1484
1485 default:
1486 fdret = FLOW_DISSECT_RET_OUT_BAD;
1487 break;
1488 }
1489
1490 /* Process result of proto processing */
1491 switch (fdret) {
1492 case FLOW_DISSECT_RET_OUT_GOOD:
1493 goto out_good;
1494 case FLOW_DISSECT_RET_PROTO_AGAIN:
1495 if (skb_flow_dissect_allowed(&num_hdrs))
1496 goto proto_again;
1497 goto out_good;
1498 case FLOW_DISSECT_RET_CONTINUE:
1499 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1500 break;
1501 case FLOW_DISSECT_RET_OUT_BAD:
1502 default:
1503 goto out_bad;
1504 }
1505
1506 ip_proto_again:
1507 fdret = FLOW_DISSECT_RET_CONTINUE;
1508
1509 switch (ip_proto) {
1510 case IPPROTO_GRE:
1511 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1512 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1513 break;
1514 }
1515
1516 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1517 target_container, data,
1518 &proto, &nhoff, &hlen, flags);
1519 break;
1520
1521 case NEXTHDR_HOP:
1522 case NEXTHDR_ROUTING:
1523 case NEXTHDR_DEST: {
1524 u8 _opthdr[2], *opthdr;
1525
1526 if (proto != htons(ETH_P_IPV6))
1527 break;
1528
1529 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1530 data, hlen, &_opthdr);
1531 if (!opthdr) {
1532 fdret = FLOW_DISSECT_RET_OUT_BAD;
1533 break;
1534 }
1535
1536 ip_proto = opthdr[0];
1537 nhoff += (opthdr[1] + 1) << 3;
1538
1539 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1540 break;
1541 }
1542 case NEXTHDR_FRAGMENT: {
1543 struct frag_hdr _fh, *fh;
1544
1545 if (proto != htons(ETH_P_IPV6))
1546 break;
1547
1548 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1549 data, hlen, &_fh);
1550
1551 if (!fh) {
1552 fdret = FLOW_DISSECT_RET_OUT_BAD;
1553 break;
1554 }
1555
1556 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1557
1558 nhoff += sizeof(_fh);
1559 ip_proto = fh->nexthdr;
1560
1561 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1562 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1563 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1564 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1565 break;
1566 }
1567 }
1568
1569 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1570 break;
1571 }
1572 case IPPROTO_IPIP:
1573 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1574 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1575 break;
1576 }
1577
1578 proto = htons(ETH_P_IP);
1579
1580 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1581 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1582 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1583 break;
1584 }
1585
1586 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1587 break;
1588
1589 case IPPROTO_IPV6:
1590 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1591 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1592 break;
1593 }
1594
1595 proto = htons(ETH_P_IPV6);
1596
1597 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1598 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1599 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1600 break;
1601 }
1602
1603 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1604 break;
1605
1606
1607 case IPPROTO_MPLS:
1608 proto = htons(ETH_P_MPLS_UC);
1609 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1610 break;
1611
1612 case IPPROTO_TCP:
1613 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1614 data, nhoff, hlen);
1615 break;
1616
1617 case IPPROTO_ICMP:
1618 case IPPROTO_ICMPV6:
1619 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1620 data, nhoff, hlen);
1621 break;
1622 case IPPROTO_L2TP:
1623 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1624 data, nhoff, hlen);
1625 break;
1626 case IPPROTO_ESP:
1627 __skb_flow_dissect_esp(skb, flow_dissector, target_container,
1628 data, nhoff, hlen);
1629 break;
1630 case IPPROTO_AH:
1631 __skb_flow_dissect_ah(skb, flow_dissector, target_container,
1632 data, nhoff, hlen);
1633 break;
1634 default:
1635 break;
1636 }
1637
1638 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1639 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1640 data, nhoff, ip_proto, hlen);
1641
1642 /* Process result of IP proto processing */
1643 switch (fdret) {
1644 case FLOW_DISSECT_RET_PROTO_AGAIN:
1645 if (skb_flow_dissect_allowed(&num_hdrs))
1646 goto proto_again;
1647 break;
1648 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1649 if (skb_flow_dissect_allowed(&num_hdrs))
1650 goto ip_proto_again;
1651 break;
1652 case FLOW_DISSECT_RET_OUT_GOOD:
1653 case FLOW_DISSECT_RET_CONTINUE:
1654 break;
1655 case FLOW_DISSECT_RET_OUT_BAD:
1656 default:
1657 goto out_bad;
1658 }
1659
1660 out_good:
1661 ret = true;
1662
1663 out:
1664 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1665 key_basic->n_proto = proto;
1666 key_basic->ip_proto = ip_proto;
1667
1668 return ret;
1669
1670 out_bad:
1671 ret = false;
1672 goto out;
1673 }
1674 EXPORT_SYMBOL(__skb_flow_dissect);
1675
1676 static siphash_aligned_key_t hashrnd;
__flow_hash_secret_init(void)1677 static __always_inline void __flow_hash_secret_init(void)
1678 {
1679 net_get_random_once(&hashrnd, sizeof(hashrnd));
1680 }
1681
flow_keys_hash_start(const struct flow_keys * flow)1682 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1683 {
1684 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1685 return &flow->FLOW_KEYS_HASH_START_FIELD;
1686 }
1687
flow_keys_hash_length(const struct flow_keys * flow)1688 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1689 {
1690 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1691
1692 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1693
1694 switch (flow->control.addr_type) {
1695 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1696 diff -= sizeof(flow->addrs.v4addrs);
1697 break;
1698 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1699 diff -= sizeof(flow->addrs.v6addrs);
1700 break;
1701 case FLOW_DISSECTOR_KEY_TIPC:
1702 diff -= sizeof(flow->addrs.tipckey);
1703 break;
1704 }
1705 return sizeof(*flow) - diff;
1706 }
1707
flow_get_u32_src(const struct flow_keys * flow)1708 __be32 flow_get_u32_src(const struct flow_keys *flow)
1709 {
1710 switch (flow->control.addr_type) {
1711 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1712 return flow->addrs.v4addrs.src;
1713 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1714 return (__force __be32)ipv6_addr_hash(
1715 &flow->addrs.v6addrs.src);
1716 case FLOW_DISSECTOR_KEY_TIPC:
1717 return flow->addrs.tipckey.key;
1718 default:
1719 return 0;
1720 }
1721 }
1722 EXPORT_SYMBOL(flow_get_u32_src);
1723
flow_get_u32_dst(const struct flow_keys * flow)1724 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1725 {
1726 switch (flow->control.addr_type) {
1727 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1728 return flow->addrs.v4addrs.dst;
1729 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1730 return (__force __be32)ipv6_addr_hash(
1731 &flow->addrs.v6addrs.dst);
1732 default:
1733 return 0;
1734 }
1735 }
1736 EXPORT_SYMBOL(flow_get_u32_dst);
1737
1738 /* Sort the source and destination IP and the ports,
1739 * to have consistent hash within the two directions
1740 */
__flow_hash_consistentify(struct flow_keys * keys)1741 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1742 {
1743 int addr_diff, i;
1744
1745 switch (keys->control.addr_type) {
1746 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1747 if ((__force u32)keys->addrs.v4addrs.dst <
1748 (__force u32)keys->addrs.v4addrs.src)
1749 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1750
1751 if ((__force u16)keys->ports.dst <
1752 (__force u16)keys->ports.src) {
1753 swap(keys->ports.src, keys->ports.dst);
1754 }
1755 break;
1756 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1757 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1758 &keys->addrs.v6addrs.src,
1759 sizeof(keys->addrs.v6addrs.dst));
1760 if (addr_diff < 0) {
1761 for (i = 0; i < 4; i++)
1762 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1763 keys->addrs.v6addrs.dst.s6_addr32[i]);
1764 }
1765 if ((__force u16)keys->ports.dst <
1766 (__force u16)keys->ports.src) {
1767 swap(keys->ports.src, keys->ports.dst);
1768 }
1769 break;
1770 }
1771 }
1772
__flow_hash_from_keys(struct flow_keys * keys,const siphash_key_t * keyval)1773 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1774 const siphash_key_t *keyval)
1775 {
1776 u32 hash;
1777
1778 __flow_hash_consistentify(keys);
1779
1780 hash = siphash(flow_keys_hash_start(keys),
1781 flow_keys_hash_length(keys), keyval);
1782 if (!hash)
1783 hash = 1;
1784
1785 return hash;
1786 }
1787
flow_hash_from_keys(struct flow_keys * keys)1788 u32 flow_hash_from_keys(struct flow_keys *keys)
1789 {
1790 __flow_hash_secret_init();
1791 return __flow_hash_from_keys(keys, &hashrnd);
1792 }
1793 EXPORT_SYMBOL(flow_hash_from_keys);
1794
___skb_get_hash(const struct sk_buff * skb,struct flow_keys * keys,const siphash_key_t * keyval)1795 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1796 struct flow_keys *keys,
1797 const siphash_key_t *keyval)
1798 {
1799 skb_flow_dissect_flow_keys(skb, keys,
1800 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1801
1802 return __flow_hash_from_keys(keys, keyval);
1803 }
1804
1805 struct _flow_keys_digest_data {
1806 __be16 n_proto;
1807 u8 ip_proto;
1808 u8 padding;
1809 __be32 ports;
1810 __be32 src;
1811 __be32 dst;
1812 };
1813
make_flow_keys_digest(struct flow_keys_digest * digest,const struct flow_keys * flow)1814 void make_flow_keys_digest(struct flow_keys_digest *digest,
1815 const struct flow_keys *flow)
1816 {
1817 struct _flow_keys_digest_data *data =
1818 (struct _flow_keys_digest_data *)digest;
1819
1820 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1821
1822 memset(digest, 0, sizeof(*digest));
1823
1824 data->n_proto = flow->basic.n_proto;
1825 data->ip_proto = flow->basic.ip_proto;
1826 data->ports = flow->ports.ports;
1827 data->src = flow->addrs.v4addrs.src;
1828 data->dst = flow->addrs.v4addrs.dst;
1829 }
1830 EXPORT_SYMBOL(make_flow_keys_digest);
1831
1832 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1833
__skb_get_hash_symmetric(const struct sk_buff * skb)1834 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1835 {
1836 struct flow_keys keys;
1837
1838 __flow_hash_secret_init();
1839
1840 memset(&keys, 0, sizeof(keys));
1841 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1842 &keys, NULL, 0, 0, 0, 0);
1843
1844 return __flow_hash_from_keys(&keys, &hashrnd);
1845 }
1846 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1847
1848 /**
1849 * __skb_get_hash: calculate a flow hash
1850 * @skb: sk_buff to calculate flow hash from
1851 *
1852 * This function calculates a flow hash based on src/dst addresses
1853 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1854 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1855 * if hash is a canonical 4-tuple hash over transport ports.
1856 */
__skb_get_hash(struct sk_buff * skb)1857 void __skb_get_hash(struct sk_buff *skb)
1858 {
1859 struct flow_keys keys;
1860 u32 hash;
1861
1862 __flow_hash_secret_init();
1863
1864 hash = ___skb_get_hash(skb, &keys, &hashrnd);
1865
1866 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1867 }
1868 EXPORT_SYMBOL(__skb_get_hash);
1869
skb_get_hash_perturb(const struct sk_buff * skb,const siphash_key_t * perturb)1870 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1871 const siphash_key_t *perturb)
1872 {
1873 struct flow_keys keys;
1874
1875 return ___skb_get_hash(skb, &keys, perturb);
1876 }
1877 EXPORT_SYMBOL(skb_get_hash_perturb);
1878
__skb_get_poff(const struct sk_buff * skb,const void * data,const struct flow_keys_basic * keys,int hlen)1879 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1880 const struct flow_keys_basic *keys, int hlen)
1881 {
1882 u32 poff = keys->control.thoff;
1883
1884 /* skip L4 headers for fragments after the first */
1885 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1886 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1887 return poff;
1888
1889 switch (keys->basic.ip_proto) {
1890 case IPPROTO_TCP: {
1891 /* access doff as u8 to avoid unaligned access */
1892 const u8 *doff;
1893 u8 _doff;
1894
1895 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1896 data, hlen, &_doff);
1897 if (!doff)
1898 return poff;
1899
1900 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1901 break;
1902 }
1903 case IPPROTO_UDP:
1904 case IPPROTO_UDPLITE:
1905 poff += sizeof(struct udphdr);
1906 break;
1907 /* For the rest, we do not really care about header
1908 * extensions at this point for now.
1909 */
1910 case IPPROTO_ICMP:
1911 poff += sizeof(struct icmphdr);
1912 break;
1913 case IPPROTO_ICMPV6:
1914 poff += sizeof(struct icmp6hdr);
1915 break;
1916 case IPPROTO_IGMP:
1917 poff += sizeof(struct igmphdr);
1918 break;
1919 case IPPROTO_DCCP:
1920 poff += sizeof(struct dccp_hdr);
1921 break;
1922 case IPPROTO_SCTP:
1923 poff += sizeof(struct sctphdr);
1924 break;
1925 }
1926
1927 return poff;
1928 }
1929
1930 /**
1931 * skb_get_poff - get the offset to the payload
1932 * @skb: sk_buff to get the payload offset from
1933 *
1934 * The function will get the offset to the payload as far as it could
1935 * be dissected. The main user is currently BPF, so that we can dynamically
1936 * truncate packets without needing to push actual payload to the user
1937 * space and can analyze headers only, instead.
1938 */
skb_get_poff(const struct sk_buff * skb)1939 u32 skb_get_poff(const struct sk_buff *skb)
1940 {
1941 struct flow_keys_basic keys;
1942
1943 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1944 NULL, 0, 0, 0, 0))
1945 return 0;
1946
1947 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1948 }
1949
__get_hash_from_flowi6(const struct flowi6 * fl6,struct flow_keys * keys)1950 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1951 {
1952 memset(keys, 0, sizeof(*keys));
1953
1954 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1955 sizeof(keys->addrs.v6addrs.src));
1956 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1957 sizeof(keys->addrs.v6addrs.dst));
1958 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1959 keys->ports.src = fl6->fl6_sport;
1960 keys->ports.dst = fl6->fl6_dport;
1961 keys->keyid.keyid = fl6->fl6_gre_key;
1962 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1963 keys->basic.ip_proto = fl6->flowi6_proto;
1964
1965 return flow_hash_from_keys(keys);
1966 }
1967 EXPORT_SYMBOL(__get_hash_from_flowi6);
1968
1969 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1970 {
1971 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1972 .offset = offsetof(struct flow_keys, control),
1973 },
1974 {
1975 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1976 .offset = offsetof(struct flow_keys, basic),
1977 },
1978 {
1979 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1980 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1981 },
1982 {
1983 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1984 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1985 },
1986 {
1987 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1988 .offset = offsetof(struct flow_keys, addrs.tipckey),
1989 },
1990 {
1991 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1992 .offset = offsetof(struct flow_keys, ports),
1993 },
1994 {
1995 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1996 .offset = offsetof(struct flow_keys, vlan),
1997 },
1998 {
1999 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2000 .offset = offsetof(struct flow_keys, tags),
2001 },
2002 {
2003 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2004 .offset = offsetof(struct flow_keys, keyid),
2005 },
2006 };
2007
2008 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2009 {
2010 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2011 .offset = offsetof(struct flow_keys, control),
2012 },
2013 {
2014 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2015 .offset = offsetof(struct flow_keys, basic),
2016 },
2017 {
2018 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2019 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2020 },
2021 {
2022 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2023 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2024 },
2025 {
2026 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2027 .offset = offsetof(struct flow_keys, ports),
2028 },
2029 };
2030
2031 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2032 {
2033 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2034 .offset = offsetof(struct flow_keys, control),
2035 },
2036 {
2037 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2038 .offset = offsetof(struct flow_keys, basic),
2039 },
2040 };
2041
2042 struct flow_dissector flow_keys_dissector __read_mostly;
2043 EXPORT_SYMBOL(flow_keys_dissector);
2044
2045 struct flow_dissector flow_keys_basic_dissector __read_mostly;
2046 EXPORT_SYMBOL(flow_keys_basic_dissector);
2047
init_default_flow_dissectors(void)2048 static int __init init_default_flow_dissectors(void)
2049 {
2050 skb_flow_dissector_init(&flow_keys_dissector,
2051 flow_keys_dissector_keys,
2052 ARRAY_SIZE(flow_keys_dissector_keys));
2053 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2054 flow_keys_dissector_symmetric_keys,
2055 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2056 skb_flow_dissector_init(&flow_keys_basic_dissector,
2057 flow_keys_basic_dissector_keys,
2058 ARRAY_SIZE(flow_keys_basic_dissector_keys));
2059 return 0;
2060 }
2061 core_initcall(init_default_flow_dissectors);
2062