projects / linux-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
Merge tag 'mm-hotfixes-stable-2025-07-11-16-16' of git://git.kernel.org/pub/scm/linux...
[linux-block.git] / net / core / flow_dissector.c
... / ...
CommitLineData
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
40static 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
46void 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}
74EXPORT_SYMBOL(skb_flow_dissector_init);
75
76#ifdef CONFIG_BPF_SYSCALL
77int 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 */
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}
140EXPORT_SYMBOL(skb_flow_get_ports);
141
142static 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 */
165void 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}
186EXPORT_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 */
191static 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
208static 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
230static 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
252static 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
276void 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}
299EXPORT_SYMBOL(skb_flow_dissect_meta);
300
301static void
302skb_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
319void
320skb_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}
363EXPORT_SYMBOL(skb_flow_dissect_ct);
364
365void
366skb_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}
495EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
496
497void 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}
512EXPORT_SYMBOL(skb_flow_dissect_hash);
513
514static enum flow_dissect_ret
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
572static enum flow_dissect_ret
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
628static enum flow_dissect_ret
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
652static enum flow_dissect_ret
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 */
793static enum flow_dissect_ret
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
825static void
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
850static void
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
882static void
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
900static void
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
923static bool skb_flow_dissect_allowed(int *num_hdrs)
924{
925 ++*num_hdrs;
926
927 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
928}
929
930static 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
1004u32 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 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
1027 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
1028 flow_keys->nhoff, hlen);
1029
1030 return result;
1031}
1032
1033static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
1034{
1035 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
1036}
1037
1038/**
1039 * __skb_flow_dissect - extract the flow_keys struct and return it
1040 * @net: associated network namespace, derived from @skb if NULL
1041 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
1042 * @flow_dissector: list of keys to dissect
1043 * @target_container: target structure to put dissected values into
1044 * @data: raw buffer pointer to the packet, if NULL use skb->data
1045 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
1046 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
1047 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
1048 * @flags: flags that control the dissection process, e.g.
1049 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
1050 *
1051 * The function will try to retrieve individual keys into target specified
1052 * by flow_dissector from either the skbuff or a raw buffer specified by the
1053 * rest parameters.
1054 *
1055 * Caller must take care of zeroing target container memory.
1056 */
1057bool __skb_flow_dissect(const struct net *net,
1058 const struct sk_buff *skb,
1059 struct flow_dissector *flow_dissector,
1060 void *target_container, const void *data,
1061 __be16 proto, int nhoff, int hlen, unsigned int flags)
1062{
1063 struct flow_dissector_key_control *key_control;
1064 struct flow_dissector_key_basic *key_basic;
1065 struct flow_dissector_key_addrs *key_addrs;
1066 struct flow_dissector_key_tags *key_tags;
1067 struct flow_dissector_key_vlan *key_vlan;
1068 enum flow_dissect_ret fdret;
1069 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
1070 bool mpls_el = false;
1071 int mpls_lse = 0;
1072 int num_hdrs = 0;
1073 u8 ip_proto = 0;
1074 bool ret;
1075
1076 if (!data) {
1077 data = skb->data;
1078 proto = skb_vlan_tag_present(skb) ?
1079 skb->vlan_proto : skb->protocol;
1080 nhoff = skb_network_offset(skb);
1081 hlen = skb_headlen(skb);
1082#if IS_ENABLED(CONFIG_NET_DSA)
1083 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
1084 proto == htons(ETH_P_XDSA))) {
1085 struct metadata_dst *md_dst = skb_metadata_dst(skb);
1086 const struct dsa_device_ops *ops;
1087 int offset = 0;
1088
1089 ops = skb->dev->dsa_ptr->tag_ops;
1090 /* Only DSA header taggers break flow dissection */
1091 if (ops->needed_headroom &&
1092 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
1093 if (ops->flow_dissect)
1094 ops->flow_dissect(skb, &proto, &offset);
1095 else
1096 dsa_tag_generic_flow_dissect(skb,
1097 &proto,
1098 &offset);
1099 hlen -= offset;
1100 nhoff += offset;
1101 }
1102 }
1103#endif
1104 }
1105
1106 /* It is ensured by skb_flow_dissector_init() that control key will
1107 * be always present.
1108 */
1109 key_control = skb_flow_dissector_target(flow_dissector,
1110 FLOW_DISSECTOR_KEY_CONTROL,
1111 target_container);
1112
1113 /* It is ensured by skb_flow_dissector_init() that basic key will
1114 * be always present.
1115 */
1116 key_basic = skb_flow_dissector_target(flow_dissector,
1117 FLOW_DISSECTOR_KEY_BASIC,
1118 target_container);
1119
1120 rcu_read_lock();
1121
1122 if (skb) {
1123 if (!net) {
1124 if (skb->dev)
1125 net = dev_net_rcu(skb->dev);
1126 else if (skb->sk)
1127 net = sock_net(skb->sk);
1128 }
1129 }
1130
1131 DEBUG_NET_WARN_ON_ONCE(!net);
1132 if (net) {
1133 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1134 struct bpf_prog_array *run_array;
1135
1136 run_array = rcu_dereference(init_net.bpf.run_array[type]);
1137 if (!run_array)
1138 run_array = rcu_dereference(net->bpf.run_array[type]);
1139
1140 if (run_array) {
1141 struct bpf_flow_keys flow_keys;
1142 struct bpf_flow_dissector ctx = {
1143 .flow_keys = &flow_keys,
1144 .data = data,
1145 .data_end = data + hlen,
1146 };
1147 __be16 n_proto = proto;
1148 struct bpf_prog *prog;
1149 u32 result;
1150
1151 if (skb) {
1152 ctx.skb = skb;
1153 /* we can't use 'proto' in the skb case
1154 * because it might be set to skb->vlan_proto
1155 * which has been pulled from the data
1156 */
1157 n_proto = skb->protocol;
1158 }
1159
1160 prog = READ_ONCE(run_array->items[0].prog);
1161 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1162 hlen, flags);
1163 if (result != BPF_FLOW_DISSECTOR_CONTINUE) {
1164 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1165 target_container);
1166 rcu_read_unlock();
1167 return result == BPF_OK;
1168 }
1169 }
1170 }
1171
1172 rcu_read_unlock();
1173
1174 if (dissector_uses_key(flow_dissector,
1175 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1176 struct ethhdr *eth = eth_hdr(skb);
1177 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1178
1179 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1180 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1181 target_container);
1182 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1183 }
1184
1185 if (dissector_uses_key(flow_dissector,
1186 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1187 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1188
1189 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1190 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1191 target_container);
1192 key_num_of_vlans->num_of_vlans = 0;
1193 }
1194
1195proto_again:
1196 fdret = FLOW_DISSECT_RET_CONTINUE;
1197
1198 switch (proto) {
1199 case htons(ETH_P_IP): {
1200 const struct iphdr *iph;
1201 struct iphdr _iph;
1202
1203 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1204 if (!iph || iph->ihl < 5) {
1205 fdret = FLOW_DISSECT_RET_OUT_BAD;
1206 break;
1207 }
1208
1209 nhoff += iph->ihl * 4;
1210
1211 ip_proto = iph->protocol;
1212
1213 if (dissector_uses_key(flow_dissector,
1214 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1215 key_addrs = skb_flow_dissector_target(flow_dissector,
1216 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1217 target_container);
1218
1219 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1220 sizeof(key_addrs->v4addrs.src));
1221 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1222 sizeof(key_addrs->v4addrs.dst));
1223 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1224 }
1225
1226 __skb_flow_dissect_ipv4(skb, flow_dissector,
1227 target_container, data, iph);
1228
1229 if (ip_is_fragment(iph)) {
1230 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1231
1232 if (iph->frag_off & htons(IP_OFFSET)) {
1233 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1234 break;
1235 } else {
1236 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1237 if (!(flags &
1238 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1239 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1240 break;
1241 }
1242 }
1243 }
1244
1245 break;
1246 }
1247 case htons(ETH_P_IPV6): {
1248 const struct ipv6hdr *iph;
1249 struct ipv6hdr _iph;
1250
1251 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1252 if (!iph) {
1253 fdret = FLOW_DISSECT_RET_OUT_BAD;
1254 break;
1255 }
1256
1257 ip_proto = iph->nexthdr;
1258 nhoff += sizeof(struct ipv6hdr);
1259
1260 if (dissector_uses_key(flow_dissector,
1261 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1262 key_addrs = skb_flow_dissector_target(flow_dissector,
1263 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1264 target_container);
1265
1266 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1267 sizeof(key_addrs->v6addrs.src));
1268 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1269 sizeof(key_addrs->v6addrs.dst));
1270 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1271 }
1272
1273 if ((dissector_uses_key(flow_dissector,
1274 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1275 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1276 ip6_flowlabel(iph)) {
1277 __be32 flow_label = ip6_flowlabel(iph);
1278
1279 if (dissector_uses_key(flow_dissector,
1280 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1281 key_tags = skb_flow_dissector_target(flow_dissector,
1282 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1283 target_container);
1284 key_tags->flow_label = ntohl(flow_label);
1285 }
1286 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1287 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1288 break;
1289 }
1290 }
1291
1292 __skb_flow_dissect_ipv6(skb, flow_dissector,
1293 target_container, data, iph);
1294
1295 break;
1296 }
1297 case htons(ETH_P_8021AD):
1298 case htons(ETH_P_8021Q): {
1299 const struct vlan_hdr *vlan = NULL;
1300 struct vlan_hdr _vlan;
1301 __be16 saved_vlan_tpid = proto;
1302
1303 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1304 skb && skb_vlan_tag_present(skb)) {
1305 proto = skb->protocol;
1306 } else {
1307 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1308 data, hlen, &_vlan);
1309 if (!vlan) {
1310 fdret = FLOW_DISSECT_RET_OUT_BAD;
1311 break;
1312 }
1313
1314 proto = vlan->h_vlan_encapsulated_proto;
1315 nhoff += sizeof(*vlan);
1316 }
1317
1318 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1319 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1320 struct flow_dissector_key_num_of_vlans *key_nvs;
1321
1322 key_nvs = skb_flow_dissector_target(flow_dissector,
1323 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1324 target_container);
1325 key_nvs->num_of_vlans++;
1326 }
1327
1328 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1329 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1330 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1331 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1332 } else {
1333 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1334 break;
1335 }
1336
1337 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1338 key_vlan = skb_flow_dissector_target(flow_dissector,
1339 dissector_vlan,
1340 target_container);
1341
1342 if (!vlan) {
1343 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1344 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1345 } else {
1346 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1347 VLAN_VID_MASK;
1348 key_vlan->vlan_priority =
1349 (ntohs(vlan->h_vlan_TCI) &
1350 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1351 }
1352 key_vlan->vlan_tpid = saved_vlan_tpid;
1353 key_vlan->vlan_eth_type = proto;
1354 }
1355
1356 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1357 break;
1358 }
1359 case htons(ETH_P_PPP_SES): {
1360 struct {
1361 struct pppoe_hdr hdr;
1362 __be16 proto;
1363 } *hdr, _hdr;
1364 u16 ppp_proto;
1365
1366 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1367 if (!hdr) {
1368 fdret = FLOW_DISSECT_RET_OUT_BAD;
1369 break;
1370 }
1371
1372 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1373 fdret = FLOW_DISSECT_RET_OUT_BAD;
1374 break;
1375 }
1376
1377 /* least significant bit of the most significant octet
1378 * indicates if protocol field was compressed
1379 */
1380 ppp_proto = ntohs(hdr->proto);
1381 if (ppp_proto & 0x0100) {
1382 ppp_proto = ppp_proto >> 8;
1383 nhoff += PPPOE_SES_HLEN - 1;
1384 } else {
1385 nhoff += PPPOE_SES_HLEN;
1386 }
1387
1388 if (ppp_proto == PPP_IP) {
1389 proto = htons(ETH_P_IP);
1390 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1391 } else if (ppp_proto == PPP_IPV6) {
1392 proto = htons(ETH_P_IPV6);
1393 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1394 } else if (ppp_proto == PPP_MPLS_UC) {
1395 proto = htons(ETH_P_MPLS_UC);
1396 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1397 } else if (ppp_proto == PPP_MPLS_MC) {
1398 proto = htons(ETH_P_MPLS_MC);
1399 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1400 } else if (ppp_proto_is_valid(ppp_proto)) {
1401 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1402 } else {
1403 fdret = FLOW_DISSECT_RET_OUT_BAD;
1404 break;
1405 }
1406
1407 if (dissector_uses_key(flow_dissector,
1408 FLOW_DISSECTOR_KEY_PPPOE)) {
1409 struct flow_dissector_key_pppoe *key_pppoe;
1410
1411 key_pppoe = skb_flow_dissector_target(flow_dissector,
1412 FLOW_DISSECTOR_KEY_PPPOE,
1413 target_container);
1414 key_pppoe->session_id = hdr->hdr.sid;
1415 key_pppoe->ppp_proto = htons(ppp_proto);
1416 key_pppoe->type = htons(ETH_P_PPP_SES);
1417 }
1418 break;
1419 }
1420 case htons(ETH_P_TIPC): {
1421 struct tipc_basic_hdr *hdr, _hdr;
1422
1423 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1424 data, hlen, &_hdr);
1425 if (!hdr) {
1426 fdret = FLOW_DISSECT_RET_OUT_BAD;
1427 break;
1428 }
1429
1430 if (dissector_uses_key(flow_dissector,
1431 FLOW_DISSECTOR_KEY_TIPC)) {
1432 key_addrs = skb_flow_dissector_target(flow_dissector,
1433 FLOW_DISSECTOR_KEY_TIPC,
1434 target_container);
1435 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1436 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1437 }
1438 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1439 break;
1440 }
1441
1442 case htons(ETH_P_MPLS_UC):
1443 case htons(ETH_P_MPLS_MC):
1444 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1445 target_container, data,
1446 nhoff, hlen, mpls_lse,
1447 &mpls_el);
1448 nhoff += sizeof(struct mpls_label);
1449 mpls_lse++;
1450 break;
1451 case htons(ETH_P_FCOE):
1452 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1453 fdret = FLOW_DISSECT_RET_OUT_BAD;
1454 break;
1455 }
1456
1457 nhoff += FCOE_HEADER_LEN;
1458 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1459 break;
1460
1461 case htons(ETH_P_ARP):
1462 case htons(ETH_P_RARP):
1463 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1464 target_container, data,
1465 nhoff, hlen);
1466 break;
1467
1468 case htons(ETH_P_BATMAN):
1469 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1470 &proto, &nhoff, hlen, flags);
1471 break;
1472
1473 case htons(ETH_P_1588): {
1474 struct ptp_header *hdr, _hdr;
1475
1476 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1477 hlen, &_hdr);
1478 if (!hdr) {
1479 fdret = FLOW_DISSECT_RET_OUT_BAD;
1480 break;
1481 }
1482
1483 nhoff += sizeof(struct ptp_header);
1484 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1485 break;
1486 }
1487
1488 case htons(ETH_P_PRP):
1489 case htons(ETH_P_HSR): {
1490 struct hsr_tag *hdr, _hdr;
1491
1492 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1493 &_hdr);
1494 if (!hdr) {
1495 fdret = FLOW_DISSECT_RET_OUT_BAD;
1496 break;
1497 }
1498
1499 proto = hdr->encap_proto;
1500 nhoff += HSR_HLEN;
1501 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1502 break;
1503 }
1504
1505 case htons(ETH_P_CFM):
1506 fdret = __skb_flow_dissect_cfm(skb, flow_dissector,
1507 target_container, data,
1508 nhoff, hlen);
1509 break;
1510
1511 default:
1512 fdret = FLOW_DISSECT_RET_OUT_BAD;
1513 break;
1514 }
1515
1516 /* Process result of proto processing */
1517 switch (fdret) {
1518 case FLOW_DISSECT_RET_OUT_GOOD:
1519 goto out_good;
1520 case FLOW_DISSECT_RET_PROTO_AGAIN:
1521 if (skb_flow_dissect_allowed(&num_hdrs))
1522 goto proto_again;
1523 goto out_good;
1524 case FLOW_DISSECT_RET_CONTINUE:
1525 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1526 break;
1527 case FLOW_DISSECT_RET_OUT_BAD:
1528 default:
1529 goto out_bad;
1530 }
1531
1532ip_proto_again:
1533 fdret = FLOW_DISSECT_RET_CONTINUE;
1534
1535 switch (ip_proto) {
1536 case IPPROTO_GRE:
1537 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1538 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1539 break;
1540 }
1541
1542 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1543 target_container, data,
1544 &proto, &nhoff, &hlen, flags);
1545 break;
1546
1547 case NEXTHDR_HOP:
1548 case NEXTHDR_ROUTING:
1549 case NEXTHDR_DEST: {
1550 u8 _opthdr[2], *opthdr;
1551
1552 if (proto != htons(ETH_P_IPV6))
1553 break;
1554
1555 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1556 data, hlen, &_opthdr);
1557 if (!opthdr) {
1558 fdret = FLOW_DISSECT_RET_OUT_BAD;
1559 break;
1560 }
1561
1562 ip_proto = opthdr[0];
1563 nhoff += (opthdr[1] + 1) << 3;
1564
1565 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1566 break;
1567 }
1568 case NEXTHDR_FRAGMENT: {
1569 struct frag_hdr _fh, *fh;
1570
1571 if (proto != htons(ETH_P_IPV6))
1572 break;
1573
1574 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1575 data, hlen, &_fh);
1576
1577 if (!fh) {
1578 fdret = FLOW_DISSECT_RET_OUT_BAD;
1579 break;
1580 }
1581
1582 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1583
1584 nhoff += sizeof(_fh);
1585 ip_proto = fh->nexthdr;
1586
1587 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1588 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1589 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1590 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1591 break;
1592 }
1593 }
1594
1595 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1596 break;
1597 }
1598 case IPPROTO_IPIP:
1599 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1600 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1601 break;
1602 }
1603
1604 proto = htons(ETH_P_IP);
1605
1606 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1607 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1608 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1609 break;
1610 }
1611
1612 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1613 break;
1614
1615 case IPPROTO_IPV6:
1616 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1617 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1618 break;
1619 }
1620
1621 proto = htons(ETH_P_IPV6);
1622
1623 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1624 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1625 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1626 break;
1627 }
1628
1629 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1630 break;
1631
1632
1633 case IPPROTO_MPLS:
1634 proto = htons(ETH_P_MPLS_UC);
1635 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1636 break;
1637
1638 case IPPROTO_TCP:
1639 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1640 data, nhoff, hlen);
1641 break;
1642
1643 case IPPROTO_ICMP:
1644 case IPPROTO_ICMPV6:
1645 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1646 data, nhoff, hlen);
1647 break;
1648 case IPPROTO_L2TP:
1649 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1650 data, nhoff, hlen);
1651 break;
1652 case IPPROTO_ESP:
1653 __skb_flow_dissect_esp(skb, flow_dissector, target_container,
1654 data, nhoff, hlen);
1655 break;
1656 case IPPROTO_AH:
1657 __skb_flow_dissect_ah(skb, flow_dissector, target_container,
1658 data, nhoff, hlen);
1659 break;
1660 default:
1661 break;
1662 }
1663
1664 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1665 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1666 data, nhoff, ip_proto, hlen);
1667
1668 /* Process result of IP proto processing */
1669 switch (fdret) {
1670 case FLOW_DISSECT_RET_PROTO_AGAIN:
1671 if (skb_flow_dissect_allowed(&num_hdrs))
1672 goto proto_again;
1673 break;
1674 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1675 if (skb_flow_dissect_allowed(&num_hdrs))
1676 goto ip_proto_again;
1677 break;
1678 case FLOW_DISSECT_RET_OUT_GOOD:
1679 case FLOW_DISSECT_RET_CONTINUE:
1680 break;
1681 case FLOW_DISSECT_RET_OUT_BAD:
1682 default:
1683 goto out_bad;
1684 }
1685
1686out_good:
1687 ret = true;
1688
1689out:
1690 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1691 key_basic->n_proto = proto;
1692 key_basic->ip_proto = ip_proto;
1693
1694 return ret;
1695
1696out_bad:
1697 ret = false;
1698 goto out;
1699}
1700EXPORT_SYMBOL(__skb_flow_dissect);
1701
1702static siphash_aligned_key_t hashrnd;
1703static __always_inline void __flow_hash_secret_init(void)
1704{
1705 net_get_random_once(&hashrnd, sizeof(hashrnd));
1706}
1707
1708static const void *flow_keys_hash_start(const struct flow_keys *flow)
1709{
1710 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1711 return &flow->FLOW_KEYS_HASH_START_FIELD;
1712}
1713
1714static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1715{
1716 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1717
1718 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1719
1720 switch (flow->control.addr_type) {
1721 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1722 diff -= sizeof(flow->addrs.v4addrs);
1723 break;
1724 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1725 diff -= sizeof(flow->addrs.v6addrs);
1726 break;
1727 case FLOW_DISSECTOR_KEY_TIPC:
1728 diff -= sizeof(flow->addrs.tipckey);
1729 break;
1730 }
1731 return sizeof(*flow) - diff;
1732}
1733
1734__be32 flow_get_u32_src(const struct flow_keys *flow)
1735{
1736 switch (flow->control.addr_type) {
1737 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1738 return flow->addrs.v4addrs.src;
1739 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1740 return (__force __be32)ipv6_addr_hash(
1741 &flow->addrs.v6addrs.src);
1742 case FLOW_DISSECTOR_KEY_TIPC:
1743 return flow->addrs.tipckey.key;
1744 default:
1745 return 0;
1746 }
1747}
1748EXPORT_SYMBOL(flow_get_u32_src);
1749
1750__be32 flow_get_u32_dst(const struct flow_keys *flow)
1751{
1752 switch (flow->control.addr_type) {
1753 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1754 return flow->addrs.v4addrs.dst;
1755 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1756 return (__force __be32)ipv6_addr_hash(
1757 &flow->addrs.v6addrs.dst);
1758 default:
1759 return 0;
1760 }
1761}
1762EXPORT_SYMBOL(flow_get_u32_dst);
1763
1764/* Sort the source and destination IP and the ports,
1765 * to have consistent hash within the two directions
1766 */
1767static inline void __flow_hash_consistentify(struct flow_keys *keys)
1768{
1769 int addr_diff, i;
1770
1771 switch (keys->control.addr_type) {
1772 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1773 if ((__force u32)keys->addrs.v4addrs.dst <
1774 (__force u32)keys->addrs.v4addrs.src)
1775 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1776
1777 if ((__force u16)keys->ports.dst <
1778 (__force u16)keys->ports.src) {
1779 swap(keys->ports.src, keys->ports.dst);
1780 }
1781 break;
1782 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1783 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1784 &keys->addrs.v6addrs.src,
1785 sizeof(keys->addrs.v6addrs.dst));
1786 if (addr_diff < 0) {
1787 for (i = 0; i < 4; i++)
1788 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1789 keys->addrs.v6addrs.dst.s6_addr32[i]);
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 }
1797}
1798
1799static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1800 const siphash_key_t *keyval)
1801{
1802 u32 hash;
1803
1804 __flow_hash_consistentify(keys);
1805
1806 hash = siphash(flow_keys_hash_start(keys),
1807 flow_keys_hash_length(keys), keyval);
1808 if (!hash)
1809 hash = 1;
1810
1811 return hash;
1812}
1813
1814u32 flow_hash_from_keys(struct flow_keys *keys)
1815{
1816 __flow_hash_secret_init();
1817 return __flow_hash_from_keys(keys, &hashrnd);
1818}
1819EXPORT_SYMBOL(flow_hash_from_keys);
1820
1821u32 flow_hash_from_keys_seed(struct flow_keys *keys,
1822 const siphash_key_t *keyval)
1823{
1824 return __flow_hash_from_keys(keys, keyval);
1825}
1826EXPORT_SYMBOL(flow_hash_from_keys_seed);
1827
1828static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1829 struct flow_keys *keys,
1830 const siphash_key_t *keyval)
1831{
1832 skb_flow_dissect_flow_keys(skb, keys,
1833 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1834
1835 return __flow_hash_from_keys(keys, keyval);
1836}
1837
1838struct _flow_keys_digest_data {
1839 __be16 n_proto;
1840 u8 ip_proto;
1841 u8 padding;
1842 __be32 ports;
1843 __be32 src;
1844 __be32 dst;
1845};
1846
1847void make_flow_keys_digest(struct flow_keys_digest *digest,
1848 const struct flow_keys *flow)
1849{
1850 struct _flow_keys_digest_data *data =
1851 (struct _flow_keys_digest_data *)digest;
1852
1853 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1854
1855 memset(digest, 0, sizeof(*digest));
1856
1857 data->n_proto = flow->basic.n_proto;
1858 data->ip_proto = flow->basic.ip_proto;
1859 data->ports = flow->ports.ports;
1860 data->src = flow->addrs.v4addrs.src;
1861 data->dst = flow->addrs.v4addrs.dst;
1862}
1863EXPORT_SYMBOL(make_flow_keys_digest);
1864
1865static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1866
1867u32 __skb_get_hash_symmetric_net(const struct net *net, const struct sk_buff *skb)
1868{
1869 struct flow_keys keys;
1870
1871 __flow_hash_secret_init();
1872
1873 memset(&keys, 0, sizeof(keys));
1874 __skb_flow_dissect(net, skb, &flow_keys_dissector_symmetric,
1875 &keys, NULL, 0, 0, 0, 0);
1876
1877 return __flow_hash_from_keys(&keys, &hashrnd);
1878}
1879EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric_net);
1880
1881/**
1882 * __skb_get_hash_net: calculate a flow hash
1883 * @net: associated network namespace, derived from @skb if NULL
1884 * @skb: sk_buff to calculate flow hash from
1885 *
1886 * This function calculates a flow hash based on src/dst addresses
1887 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1888 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1889 * if hash is a canonical 4-tuple hash over transport ports.
1890 */
1891void __skb_get_hash_net(const struct net *net, struct sk_buff *skb)
1892{
1893 struct flow_keys keys;
1894 u32 hash;
1895
1896 memset(&keys, 0, sizeof(keys));
1897
1898 __skb_flow_dissect(net, skb, &flow_keys_dissector,
1899 &keys, NULL, 0, 0, 0,
1900 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1901
1902 __flow_hash_secret_init();
1903
1904 hash = __flow_hash_from_keys(&keys, &hashrnd);
1905
1906 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1907}
1908EXPORT_SYMBOL(__skb_get_hash_net);
1909
1910__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1911 const siphash_key_t *perturb)
1912{
1913 struct flow_keys keys;
1914
1915 return ___skb_get_hash(skb, &keys, perturb);
1916}
1917EXPORT_SYMBOL(skb_get_hash_perturb);
1918
1919u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1920 const struct flow_keys_basic *keys, int hlen)
1921{
1922 u32 poff = keys->control.thoff;
1923
1924 /* skip L4 headers for fragments after the first */
1925 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1926 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1927 return poff;
1928
1929 switch (keys->basic.ip_proto) {
1930 case IPPROTO_TCP: {
1931 /* access doff as u8 to avoid unaligned access */
1932 const u8 *doff;
1933 u8 _doff;
1934
1935 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1936 data, hlen, &_doff);
1937 if (!doff)
1938 return poff;
1939
1940 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1941 break;
1942 }
1943 case IPPROTO_UDP:
1944 case IPPROTO_UDPLITE:
1945 poff += sizeof(struct udphdr);
1946 break;
1947 /* For the rest, we do not really care about header
1948 * extensions at this point for now.
1949 */
1950 case IPPROTO_ICMP:
1951 poff += sizeof(struct icmphdr);
1952 break;
1953 case IPPROTO_ICMPV6:
1954 poff += sizeof(struct icmp6hdr);
1955 break;
1956 case IPPROTO_IGMP:
1957 poff += sizeof(struct igmphdr);
1958 break;
1959 case IPPROTO_DCCP:
1960 poff += sizeof(struct dccp_hdr);
1961 break;
1962 case IPPROTO_SCTP:
1963 poff += sizeof(struct sctphdr);
1964 break;
1965 }
1966
1967 return poff;
1968}
1969
1970/**
1971 * skb_get_poff - get the offset to the payload
1972 * @skb: sk_buff to get the payload offset from
1973 *
1974 * The function will get the offset to the payload as far as it could
1975 * be dissected. The main user is currently BPF, so that we can dynamically
1976 * truncate packets without needing to push actual payload to the user
1977 * space and can analyze headers only, instead.
1978 */
1979u32 skb_get_poff(const struct sk_buff *skb)
1980{
1981 struct flow_keys_basic keys;
1982
1983 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1984 NULL, 0, 0, 0, 0))
1985 return 0;
1986
1987 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1988}
1989
1990__u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1991{
1992 memset(keys, 0, sizeof(*keys));
1993
1994 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1995 sizeof(keys->addrs.v6addrs.src));
1996 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1997 sizeof(keys->addrs.v6addrs.dst));
1998 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1999 keys->ports.src = fl6->fl6_sport;
2000 keys->ports.dst = fl6->fl6_dport;
2001 keys->keyid.keyid = fl6->fl6_gre_key;
2002 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2003 keys->basic.ip_proto = fl6->flowi6_proto;
2004
2005 return flow_hash_from_keys(keys);
2006}
2007EXPORT_SYMBOL(__get_hash_from_flowi6);
2008
2009static const struct flow_dissector_key flow_keys_dissector_keys[] = {
2010 {
2011 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2012 .offset = offsetof(struct flow_keys, control),
2013 },
2014 {
2015 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2016 .offset = offsetof(struct flow_keys, basic),
2017 },
2018 {
2019 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2020 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2021 },
2022 {
2023 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2024 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2025 },
2026 {
2027 .key_id = FLOW_DISSECTOR_KEY_TIPC,
2028 .offset = offsetof(struct flow_keys, addrs.tipckey),
2029 },
2030 {
2031 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2032 .offset = offsetof(struct flow_keys, ports),
2033 },
2034 {
2035 .key_id = FLOW_DISSECTOR_KEY_VLAN,
2036 .offset = offsetof(struct flow_keys, vlan),
2037 },
2038 {
2039 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
2040 .offset = offsetof(struct flow_keys, tags),
2041 },
2042 {
2043 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
2044 .offset = offsetof(struct flow_keys, keyid),
2045 },
2046};
2047
2048static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
2049 {
2050 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2051 .offset = offsetof(struct flow_keys, control),
2052 },
2053 {
2054 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2055 .offset = offsetof(struct flow_keys, basic),
2056 },
2057 {
2058 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
2059 .offset = offsetof(struct flow_keys, addrs.v4addrs),
2060 },
2061 {
2062 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
2063 .offset = offsetof(struct flow_keys, addrs.v6addrs),
2064 },
2065 {
2066 .key_id = FLOW_DISSECTOR_KEY_PORTS,
2067 .offset = offsetof(struct flow_keys, ports),
2068 },
2069};
2070
2071static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
2072 {
2073 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
2074 .offset = offsetof(struct flow_keys, control),
2075 },
2076 {
2077 .key_id = FLOW_DISSECTOR_KEY_BASIC,
2078 .offset = offsetof(struct flow_keys, basic),
2079 },
2080};
2081
2082struct flow_dissector flow_keys_dissector __read_mostly;
2083EXPORT_SYMBOL(flow_keys_dissector);
2084
2085struct flow_dissector flow_keys_basic_dissector __read_mostly;
2086EXPORT_SYMBOL(flow_keys_basic_dissector);
2087
2088static int __init init_default_flow_dissectors(void)
2089{
2090 skb_flow_dissector_init(&flow_keys_dissector,
2091 flow_keys_dissector_keys,
2092 ARRAY_SIZE(flow_keys_dissector_keys));
2093 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
2094 flow_keys_dissector_symmetric_keys,
2095 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
2096 skb_flow_dissector_init(&flow_keys_basic_dissector,
2097 flow_keys_basic_dissector_keys,
2098 ARRAY_SIZE(flow_keys_basic_dissector_keys));
2099 return 0;
2100}
2101core_initcall(init_default_flow_dissectors);
Linux 6.x block layer and io_uring tree(s)