1 // SPDX-License-Identifier: GPL-2.0-only
2 #include <linux/kernel.h>
3 #include <linux/skbuff.h>
4 #include <linux/export.h>
6 #include <linux/ipv6.h>
7 #include <linux/if_vlan.h>
9 #include <net/dst_metadata.h>
15 #include <linux/igmp.h>
16 #include <linux/icmp.h>
17 #include <linux/sctp.h>
18 #include <linux/dccp.h>
19 #include <linux/if_tunnel.h>
20 #include <linux/if_pppox.h>
21 #include <linux/ppp_defs.h>
22 #include <linux/stddef.h>
23 #include <linux/if_ether.h>
24 #include <linux/mpls.h>
25 #include <linux/tcp.h>
26 #include <net/flow_dissector.h>
27 #include <scsi/fc/fc_fcoe.h>
28 #include <uapi/linux/batadv_packet.h>
29 #include <linux/bpf.h>
30 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
31 #include <net/netfilter/nf_conntrack_core.h>
32 #include <net/netfilter/nf_conntrack_labels.h>
35 static DEFINE_MUTEX(flow_dissector_mutex);
37 static void dissector_set_key(struct flow_dissector *flow_dissector,
38 enum flow_dissector_key_id key_id)
40 flow_dissector->used_keys |= (1 << key_id);
43 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
44 const struct flow_dissector_key *key,
45 unsigned int key_count)
49 memset(flow_dissector, 0, sizeof(*flow_dissector));
51 for (i = 0; i < key_count; i++, key++) {
52 /* User should make sure that every key target offset is withing
53 * boundaries of unsigned short.
55 BUG_ON(key->offset > USHRT_MAX);
56 BUG_ON(dissector_uses_key(flow_dissector,
59 dissector_set_key(flow_dissector, key->key_id);
60 flow_dissector->offset[key->key_id] = key->offset;
63 /* Ensure that the dissector always includes control and basic key.
64 * That way we are able to avoid handling lack of these in fast path.
66 BUG_ON(!dissector_uses_key(flow_dissector,
67 FLOW_DISSECTOR_KEY_CONTROL));
68 BUG_ON(!dissector_uses_key(flow_dissector,
69 FLOW_DISSECTOR_KEY_BASIC));
71 EXPORT_SYMBOL(skb_flow_dissector_init);
73 int skb_flow_dissector_prog_query(const union bpf_attr *attr,
74 union bpf_attr __user *uattr)
76 __u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids);
77 u32 prog_id, prog_cnt = 0, flags = 0;
78 struct bpf_prog *attached;
81 if (attr->query.query_flags)
84 net = get_net_ns_by_fd(attr->query.target_fd);
89 attached = rcu_dereference(net->flow_dissector_prog);
92 prog_id = attached->aux->id;
98 if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
100 if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt)))
103 if (!attr->query.prog_cnt || !prog_ids || !prog_cnt)
106 if (copy_to_user(prog_ids, &prog_id, sizeof(u32)))
112 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
113 struct bpf_prog *prog)
115 struct bpf_prog *attached;
118 net = current->nsproxy->net_ns;
119 mutex_lock(&flow_dissector_mutex);
120 attached = rcu_dereference_protected(net->flow_dissector_prog,
121 lockdep_is_held(&flow_dissector_mutex));
123 /* Only one BPF program can be attached at a time */
124 mutex_unlock(&flow_dissector_mutex);
127 rcu_assign_pointer(net->flow_dissector_prog, prog);
128 mutex_unlock(&flow_dissector_mutex);
132 static int flow_dissector_bpf_prog_detach(struct net *net)
134 struct bpf_prog *attached;
136 mutex_lock(&flow_dissector_mutex);
137 attached = rcu_dereference_protected(net->flow_dissector_prog,
138 lockdep_is_held(&flow_dissector_mutex));
140 mutex_unlock(&flow_dissector_mutex);
143 RCU_INIT_POINTER(net->flow_dissector_prog, NULL);
144 bpf_prog_put(attached);
145 mutex_unlock(&flow_dissector_mutex);
149 * skb_flow_get_be16 - extract be16 entity
150 * @skb: sk_buff to extract from
151 * @poff: offset to extract at
152 * @data: raw buffer pointer to the packet
153 * @hlen: packet header length
155 * The function will try to retrieve a be32 entity at
158 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
159 void *data, int hlen)
163 u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
170 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
172 return flow_dissector_bpf_prog_detach(current->nsproxy->net_ns);
175 static void __net_exit flow_dissector_pernet_pre_exit(struct net *net)
177 /* We're not racing with attach/detach because there are no
178 * references to netns left when pre_exit gets called.
180 if (rcu_access_pointer(net->flow_dissector_prog))
181 flow_dissector_bpf_prog_detach(net);
184 static struct pernet_operations flow_dissector_pernet_ops __net_initdata = {
185 .pre_exit = flow_dissector_pernet_pre_exit,
189 * __skb_flow_get_ports - extract the upper layer ports and return them
190 * @skb: sk_buff to extract the ports from
191 * @thoff: transport header offset
192 * @ip_proto: protocol for which to get port offset
193 * @data: raw buffer pointer to the packet, if NULL use skb->data
194 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
196 * The function will try to retrieve the ports at offset thoff + poff where poff
197 * is the protocol port offset returned from proto_ports_offset
199 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
200 void *data, int hlen)
202 int poff = proto_ports_offset(ip_proto);
206 hlen = skb_headlen(skb);
210 __be32 *ports, _ports;
212 ports = __skb_header_pointer(skb, thoff + poff,
213 sizeof(_ports), data, hlen, &_ports);
220 EXPORT_SYMBOL(__skb_flow_get_ports);
222 void skb_flow_dissect_meta(const struct sk_buff *skb,
223 struct flow_dissector *flow_dissector,
224 void *target_container)
226 struct flow_dissector_key_meta *meta;
228 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
231 meta = skb_flow_dissector_target(flow_dissector,
232 FLOW_DISSECTOR_KEY_META,
234 meta->ingress_ifindex = skb->skb_iif;
236 EXPORT_SYMBOL(skb_flow_dissect_meta);
239 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
240 struct flow_dissector *flow_dissector,
241 void *target_container)
243 struct flow_dissector_key_control *ctrl;
245 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
248 ctrl = skb_flow_dissector_target(flow_dissector,
249 FLOW_DISSECTOR_KEY_ENC_CONTROL,
251 ctrl->addr_type = type;
255 skb_flow_dissect_ct(const struct sk_buff *skb,
256 struct flow_dissector *flow_dissector,
257 void *target_container,
261 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
262 struct flow_dissector_key_ct *key;
263 enum ip_conntrack_info ctinfo;
264 struct nf_conn_labels *cl;
267 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
270 ct = nf_ct_get(skb, &ctinfo);
274 key = skb_flow_dissector_target(flow_dissector,
275 FLOW_DISSECTOR_KEY_CT,
278 if (ctinfo < mapsize)
279 key->ct_state = ctinfo_map[ctinfo];
280 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
281 key->ct_zone = ct->zone.id;
283 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
284 key->ct_mark = ct->mark;
287 cl = nf_ct_labels_find(ct);
289 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
290 #endif /* CONFIG_NF_CONNTRACK */
292 EXPORT_SYMBOL(skb_flow_dissect_ct);
295 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
296 struct flow_dissector *flow_dissector,
297 void *target_container)
299 struct ip_tunnel_info *info;
300 struct ip_tunnel_key *key;
302 /* A quick check to see if there might be something to do. */
303 if (!dissector_uses_key(flow_dissector,
304 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
305 !dissector_uses_key(flow_dissector,
306 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
307 !dissector_uses_key(flow_dissector,
308 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
309 !dissector_uses_key(flow_dissector,
310 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
311 !dissector_uses_key(flow_dissector,
312 FLOW_DISSECTOR_KEY_ENC_PORTS) &&
313 !dissector_uses_key(flow_dissector,
314 FLOW_DISSECTOR_KEY_ENC_IP) &&
315 !dissector_uses_key(flow_dissector,
316 FLOW_DISSECTOR_KEY_ENC_OPTS))
319 info = skb_tunnel_info(skb);
325 switch (ip_tunnel_info_af(info)) {
327 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
330 if (dissector_uses_key(flow_dissector,
331 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
332 struct flow_dissector_key_ipv4_addrs *ipv4;
334 ipv4 = skb_flow_dissector_target(flow_dissector,
335 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
337 ipv4->src = key->u.ipv4.src;
338 ipv4->dst = key->u.ipv4.dst;
342 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
345 if (dissector_uses_key(flow_dissector,
346 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
347 struct flow_dissector_key_ipv6_addrs *ipv6;
349 ipv6 = skb_flow_dissector_target(flow_dissector,
350 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
352 ipv6->src = key->u.ipv6.src;
353 ipv6->dst = key->u.ipv6.dst;
358 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
359 struct flow_dissector_key_keyid *keyid;
361 keyid = skb_flow_dissector_target(flow_dissector,
362 FLOW_DISSECTOR_KEY_ENC_KEYID,
364 keyid->keyid = tunnel_id_to_key32(key->tun_id);
367 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
368 struct flow_dissector_key_ports *tp;
370 tp = skb_flow_dissector_target(flow_dissector,
371 FLOW_DISSECTOR_KEY_ENC_PORTS,
373 tp->src = key->tp_src;
374 tp->dst = key->tp_dst;
377 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
378 struct flow_dissector_key_ip *ip;
380 ip = skb_flow_dissector_target(flow_dissector,
381 FLOW_DISSECTOR_KEY_ENC_IP,
387 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
388 struct flow_dissector_key_enc_opts *enc_opt;
390 enc_opt = skb_flow_dissector_target(flow_dissector,
391 FLOW_DISSECTOR_KEY_ENC_OPTS,
394 if (info->options_len) {
395 enc_opt->len = info->options_len;
396 ip_tunnel_info_opts_get(enc_opt->data, info);
397 enc_opt->dst_opt_type = info->key.tun_flags &
398 TUNNEL_OPTIONS_PRESENT;
402 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
404 static enum flow_dissect_ret
405 __skb_flow_dissect_mpls(const struct sk_buff *skb,
406 struct flow_dissector *flow_dissector,
407 void *target_container, void *data, int nhoff, int hlen)
409 struct flow_dissector_key_keyid *key_keyid;
410 struct mpls_label *hdr, _hdr[2];
413 if (!dissector_uses_key(flow_dissector,
414 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
415 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
416 return FLOW_DISSECT_RET_OUT_GOOD;
418 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
421 return FLOW_DISSECT_RET_OUT_BAD;
423 entry = ntohl(hdr[0].entry);
424 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
426 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
427 struct flow_dissector_key_mpls *key_mpls;
429 key_mpls = skb_flow_dissector_target(flow_dissector,
430 FLOW_DISSECTOR_KEY_MPLS,
432 key_mpls->mpls_label = label;
433 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
434 >> MPLS_LS_TTL_SHIFT;
435 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
437 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
441 if (label == MPLS_LABEL_ENTROPY) {
442 key_keyid = skb_flow_dissector_target(flow_dissector,
443 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
445 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
447 return FLOW_DISSECT_RET_OUT_GOOD;
450 static enum flow_dissect_ret
451 __skb_flow_dissect_arp(const struct sk_buff *skb,
452 struct flow_dissector *flow_dissector,
453 void *target_container, void *data, int nhoff, int hlen)
455 struct flow_dissector_key_arp *key_arp;
457 unsigned char ar_sha[ETH_ALEN];
458 unsigned char ar_sip[4];
459 unsigned char ar_tha[ETH_ALEN];
460 unsigned char ar_tip[4];
461 } *arp_eth, _arp_eth;
462 const struct arphdr *arp;
465 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
466 return FLOW_DISSECT_RET_OUT_GOOD;
468 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
471 return FLOW_DISSECT_RET_OUT_BAD;
473 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
474 arp->ar_pro != htons(ETH_P_IP) ||
475 arp->ar_hln != ETH_ALEN ||
477 (arp->ar_op != htons(ARPOP_REPLY) &&
478 arp->ar_op != htons(ARPOP_REQUEST)))
479 return FLOW_DISSECT_RET_OUT_BAD;
481 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
482 sizeof(_arp_eth), data,
485 return FLOW_DISSECT_RET_OUT_BAD;
487 key_arp = skb_flow_dissector_target(flow_dissector,
488 FLOW_DISSECTOR_KEY_ARP,
491 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
492 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
494 /* Only store the lower byte of the opcode;
495 * this covers ARPOP_REPLY and ARPOP_REQUEST.
497 key_arp->op = ntohs(arp->ar_op) & 0xff;
499 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
500 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
502 return FLOW_DISSECT_RET_OUT_GOOD;
505 static enum flow_dissect_ret
506 __skb_flow_dissect_gre(const struct sk_buff *skb,
507 struct flow_dissector_key_control *key_control,
508 struct flow_dissector *flow_dissector,
509 void *target_container, void *data,
510 __be16 *p_proto, int *p_nhoff, int *p_hlen,
513 struct flow_dissector_key_keyid *key_keyid;
514 struct gre_base_hdr *hdr, _hdr;
518 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
519 data, *p_hlen, &_hdr);
521 return FLOW_DISSECT_RET_OUT_BAD;
523 /* Only look inside GRE without routing */
524 if (hdr->flags & GRE_ROUTING)
525 return FLOW_DISSECT_RET_OUT_GOOD;
527 /* Only look inside GRE for version 0 and 1 */
528 gre_ver = ntohs(hdr->flags & GRE_VERSION);
530 return FLOW_DISSECT_RET_OUT_GOOD;
532 *p_proto = hdr->protocol;
534 /* Version1 must be PPTP, and check the flags */
535 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
536 return FLOW_DISSECT_RET_OUT_GOOD;
539 offset += sizeof(struct gre_base_hdr);
541 if (hdr->flags & GRE_CSUM)
542 offset += FIELD_SIZEOF(struct gre_full_hdr, csum) +
543 FIELD_SIZEOF(struct gre_full_hdr, reserved1);
545 if (hdr->flags & GRE_KEY) {
549 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
551 data, *p_hlen, &_keyid);
553 return FLOW_DISSECT_RET_OUT_BAD;
555 if (dissector_uses_key(flow_dissector,
556 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
557 key_keyid = skb_flow_dissector_target(flow_dissector,
558 FLOW_DISSECTOR_KEY_GRE_KEYID,
561 key_keyid->keyid = *keyid;
563 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
565 offset += FIELD_SIZEOF(struct gre_full_hdr, key);
568 if (hdr->flags & GRE_SEQ)
569 offset += FIELD_SIZEOF(struct pptp_gre_header, seq);
572 if (*p_proto == htons(ETH_P_TEB)) {
573 const struct ethhdr *eth;
576 eth = __skb_header_pointer(skb, *p_nhoff + offset,
578 data, *p_hlen, &_eth);
580 return FLOW_DISSECT_RET_OUT_BAD;
581 *p_proto = eth->h_proto;
582 offset += sizeof(*eth);
584 /* Cap headers that we access via pointers at the
585 * end of the Ethernet header as our maximum alignment
586 * at that point is only 2 bytes.
589 *p_hlen = *p_nhoff + offset;
591 } else { /* version 1, must be PPTP */
592 u8 _ppp_hdr[PPP_HDRLEN];
595 if (hdr->flags & GRE_ACK)
596 offset += FIELD_SIZEOF(struct pptp_gre_header, ack);
598 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
600 data, *p_hlen, _ppp_hdr);
602 return FLOW_DISSECT_RET_OUT_BAD;
604 switch (PPP_PROTOCOL(ppp_hdr)) {
606 *p_proto = htons(ETH_P_IP);
609 *p_proto = htons(ETH_P_IPV6);
612 /* Could probably catch some more like MPLS */
616 offset += PPP_HDRLEN;
620 key_control->flags |= FLOW_DIS_ENCAPSULATION;
621 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
622 return FLOW_DISSECT_RET_OUT_GOOD;
624 return FLOW_DISSECT_RET_PROTO_AGAIN;
628 * __skb_flow_dissect_batadv() - dissect batman-adv header
629 * @skb: sk_buff to with the batman-adv header
630 * @key_control: flow dissectors control key
631 * @data: raw buffer pointer to the packet, if NULL use skb->data
632 * @p_proto: pointer used to update the protocol to process next
633 * @p_nhoff: pointer used to update inner network header offset
634 * @hlen: packet header length
635 * @flags: any combination of FLOW_DISSECTOR_F_*
637 * ETH_P_BATMAN packets are tried to be dissected. Only
638 * &struct batadv_unicast packets are actually processed because they contain an
639 * inner ethernet header and are usually followed by actual network header. This
640 * allows the flow dissector to continue processing the packet.
642 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
643 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
644 * otherwise FLOW_DISSECT_RET_OUT_BAD
646 static enum flow_dissect_ret
647 __skb_flow_dissect_batadv(const struct sk_buff *skb,
648 struct flow_dissector_key_control *key_control,
649 void *data, __be16 *p_proto, int *p_nhoff, int hlen,
653 struct batadv_unicast_packet batadv_unicast;
657 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
660 return FLOW_DISSECT_RET_OUT_BAD;
662 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
663 return FLOW_DISSECT_RET_OUT_BAD;
665 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
666 return FLOW_DISSECT_RET_OUT_BAD;
668 *p_proto = hdr->eth.h_proto;
669 *p_nhoff += sizeof(*hdr);
671 key_control->flags |= FLOW_DIS_ENCAPSULATION;
672 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
673 return FLOW_DISSECT_RET_OUT_GOOD;
675 return FLOW_DISSECT_RET_PROTO_AGAIN;
679 __skb_flow_dissect_tcp(const struct sk_buff *skb,
680 struct flow_dissector *flow_dissector,
681 void *target_container, void *data, int thoff, int hlen)
683 struct flow_dissector_key_tcp *key_tcp;
684 struct tcphdr *th, _th;
686 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
689 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
693 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
696 key_tcp = skb_flow_dissector_target(flow_dissector,
697 FLOW_DISSECTOR_KEY_TCP,
699 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
703 __skb_flow_dissect_ports(const struct sk_buff *skb,
704 struct flow_dissector *flow_dissector,
705 void *target_container, void *data, int nhoff,
706 u8 ip_proto, int hlen)
708 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
709 struct flow_dissector_key_ports *key_ports;
711 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
712 dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
713 else if (dissector_uses_key(flow_dissector,
714 FLOW_DISSECTOR_KEY_PORTS_RANGE))
715 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
717 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
720 key_ports = skb_flow_dissector_target(flow_dissector,
723 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
728 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
729 struct flow_dissector *flow_dissector,
730 void *target_container, void *data, const struct iphdr *iph)
732 struct flow_dissector_key_ip *key_ip;
734 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
737 key_ip = skb_flow_dissector_target(flow_dissector,
738 FLOW_DISSECTOR_KEY_IP,
740 key_ip->tos = iph->tos;
741 key_ip->ttl = iph->ttl;
745 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
746 struct flow_dissector *flow_dissector,
747 void *target_container, void *data, const struct ipv6hdr *iph)
749 struct flow_dissector_key_ip *key_ip;
751 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
754 key_ip = skb_flow_dissector_target(flow_dissector,
755 FLOW_DISSECTOR_KEY_IP,
757 key_ip->tos = ipv6_get_dsfield(iph);
758 key_ip->ttl = iph->hop_limit;
761 /* Maximum number of protocol headers that can be parsed in
764 #define MAX_FLOW_DISSECT_HDRS 15
766 static bool skb_flow_dissect_allowed(int *num_hdrs)
770 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
773 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
774 struct flow_dissector *flow_dissector,
775 void *target_container)
777 struct flow_dissector_key_ports *key_ports = NULL;
778 struct flow_dissector_key_control *key_control;
779 struct flow_dissector_key_basic *key_basic;
780 struct flow_dissector_key_addrs *key_addrs;
781 struct flow_dissector_key_tags *key_tags;
783 key_control = skb_flow_dissector_target(flow_dissector,
784 FLOW_DISSECTOR_KEY_CONTROL,
786 key_control->thoff = flow_keys->thoff;
787 if (flow_keys->is_frag)
788 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
789 if (flow_keys->is_first_frag)
790 key_control->flags |= FLOW_DIS_FIRST_FRAG;
791 if (flow_keys->is_encap)
792 key_control->flags |= FLOW_DIS_ENCAPSULATION;
794 key_basic = skb_flow_dissector_target(flow_dissector,
795 FLOW_DISSECTOR_KEY_BASIC,
797 key_basic->n_proto = flow_keys->n_proto;
798 key_basic->ip_proto = flow_keys->ip_proto;
800 if (flow_keys->addr_proto == ETH_P_IP &&
801 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
802 key_addrs = skb_flow_dissector_target(flow_dissector,
803 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
805 key_addrs->v4addrs.src = flow_keys->ipv4_src;
806 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
807 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
808 } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
809 dissector_uses_key(flow_dissector,
810 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
811 key_addrs = skb_flow_dissector_target(flow_dissector,
812 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
814 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
815 sizeof(key_addrs->v6addrs.src));
816 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
817 sizeof(key_addrs->v6addrs.dst));
818 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
821 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
822 key_ports = skb_flow_dissector_target(flow_dissector,
823 FLOW_DISSECTOR_KEY_PORTS,
825 else if (dissector_uses_key(flow_dissector,
826 FLOW_DISSECTOR_KEY_PORTS_RANGE))
827 key_ports = skb_flow_dissector_target(flow_dissector,
828 FLOW_DISSECTOR_KEY_PORTS_RANGE,
832 key_ports->src = flow_keys->sport;
833 key_ports->dst = flow_keys->dport;
836 if (dissector_uses_key(flow_dissector,
837 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
838 key_tags = skb_flow_dissector_target(flow_dissector,
839 FLOW_DISSECTOR_KEY_FLOW_LABEL,
841 key_tags->flow_label = ntohl(flow_keys->flow_label);
845 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
846 __be16 proto, int nhoff, int hlen, unsigned int flags)
848 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
851 /* Pass parameters to the BPF program */
852 memset(flow_keys, 0, sizeof(*flow_keys));
853 flow_keys->n_proto = proto;
854 flow_keys->nhoff = nhoff;
855 flow_keys->thoff = flow_keys->nhoff;
857 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
858 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
859 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
860 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
861 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
862 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
863 flow_keys->flags = flags;
866 result = BPF_PROG_RUN(prog, ctx);
869 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
870 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
871 flow_keys->nhoff, hlen);
873 return result == BPF_OK;
877 * __skb_flow_dissect - extract the flow_keys struct and return it
878 * @net: associated network namespace, derived from @skb if NULL
879 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
880 * @flow_dissector: list of keys to dissect
881 * @target_container: target structure to put dissected values into
882 * @data: raw buffer pointer to the packet, if NULL use skb->data
883 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
884 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
885 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
886 * @flags: flags that control the dissection process, e.g.
887 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
889 * The function will try to retrieve individual keys into target specified
890 * by flow_dissector from either the skbuff or a raw buffer specified by the
893 * Caller must take care of zeroing target container memory.
895 bool __skb_flow_dissect(const struct net *net,
896 const struct sk_buff *skb,
897 struct flow_dissector *flow_dissector,
898 void *target_container,
899 void *data, __be16 proto, int nhoff, int hlen,
902 struct flow_dissector_key_control *key_control;
903 struct flow_dissector_key_basic *key_basic;
904 struct flow_dissector_key_addrs *key_addrs;
905 struct flow_dissector_key_icmp *key_icmp;
906 struct flow_dissector_key_tags *key_tags;
907 struct flow_dissector_key_vlan *key_vlan;
908 struct bpf_prog *attached = NULL;
909 enum flow_dissect_ret fdret;
910 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
917 proto = skb_vlan_tag_present(skb) ?
918 skb->vlan_proto : skb->protocol;
919 nhoff = skb_network_offset(skb);
920 hlen = skb_headlen(skb);
921 #if IS_ENABLED(CONFIG_NET_DSA)
922 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
923 proto == htons(ETH_P_XDSA))) {
924 const struct dsa_device_ops *ops;
927 ops = skb->dev->dsa_ptr->tag_ops;
928 if (ops->flow_dissect &&
929 !ops->flow_dissect(skb, &proto, &offset)) {
937 /* It is ensured by skb_flow_dissector_init() that control key will
940 key_control = skb_flow_dissector_target(flow_dissector,
941 FLOW_DISSECTOR_KEY_CONTROL,
944 /* It is ensured by skb_flow_dissector_init() that basic key will
947 key_basic = skb_flow_dissector_target(flow_dissector,
948 FLOW_DISSECTOR_KEY_BASIC,
954 net = dev_net(skb->dev);
956 net = sock_net(skb->sk);
963 attached = rcu_dereference(net->flow_dissector_prog);
966 struct bpf_flow_keys flow_keys;
967 struct bpf_flow_dissector ctx = {
968 .flow_keys = &flow_keys,
970 .data_end = data + hlen,
972 __be16 n_proto = proto;
976 /* we can't use 'proto' in the skb case
977 * because it might be set to skb->vlan_proto
978 * which has been pulled from the data
980 n_proto = skb->protocol;
983 ret = bpf_flow_dissect(attached, &ctx, n_proto, nhoff,
985 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
993 if (dissector_uses_key(flow_dissector,
994 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
995 struct ethhdr *eth = eth_hdr(skb);
996 struct flow_dissector_key_eth_addrs *key_eth_addrs;
998 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
999 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1001 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs));
1005 fdret = FLOW_DISSECT_RET_CONTINUE;
1008 case htons(ETH_P_IP): {
1009 const struct iphdr *iph;
1012 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1013 if (!iph || iph->ihl < 5) {
1014 fdret = FLOW_DISSECT_RET_OUT_BAD;
1018 nhoff += iph->ihl * 4;
1020 ip_proto = iph->protocol;
1022 if (dissector_uses_key(flow_dissector,
1023 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1024 key_addrs = skb_flow_dissector_target(flow_dissector,
1025 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1028 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1029 sizeof(key_addrs->v4addrs.src));
1030 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1031 sizeof(key_addrs->v4addrs.dst));
1032 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1035 __skb_flow_dissect_ipv4(skb, flow_dissector,
1036 target_container, data, iph);
1038 if (ip_is_fragment(iph)) {
1039 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1041 if (iph->frag_off & htons(IP_OFFSET)) {
1042 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1045 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1047 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1048 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1056 case htons(ETH_P_IPV6): {
1057 const struct ipv6hdr *iph;
1058 struct ipv6hdr _iph;
1060 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1062 fdret = FLOW_DISSECT_RET_OUT_BAD;
1066 ip_proto = iph->nexthdr;
1067 nhoff += sizeof(struct ipv6hdr);
1069 if (dissector_uses_key(flow_dissector,
1070 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1071 key_addrs = skb_flow_dissector_target(flow_dissector,
1072 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1075 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1076 sizeof(key_addrs->v6addrs.src));
1077 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1078 sizeof(key_addrs->v6addrs.dst));
1079 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1082 if ((dissector_uses_key(flow_dissector,
1083 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1084 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1085 ip6_flowlabel(iph)) {
1086 __be32 flow_label = ip6_flowlabel(iph);
1088 if (dissector_uses_key(flow_dissector,
1089 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1090 key_tags = skb_flow_dissector_target(flow_dissector,
1091 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1093 key_tags->flow_label = ntohl(flow_label);
1095 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1096 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1101 __skb_flow_dissect_ipv6(skb, flow_dissector,
1102 target_container, data, iph);
1106 case htons(ETH_P_8021AD):
1107 case htons(ETH_P_8021Q): {
1108 const struct vlan_hdr *vlan = NULL;
1109 struct vlan_hdr _vlan;
1110 __be16 saved_vlan_tpid = proto;
1112 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1113 skb && skb_vlan_tag_present(skb)) {
1114 proto = skb->protocol;
1116 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1117 data, hlen, &_vlan);
1119 fdret = FLOW_DISSECT_RET_OUT_BAD;
1123 proto = vlan->h_vlan_encapsulated_proto;
1124 nhoff += sizeof(*vlan);
1127 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1128 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1129 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1130 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1132 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1136 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1137 key_vlan = skb_flow_dissector_target(flow_dissector,
1142 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1143 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1145 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1147 key_vlan->vlan_priority =
1148 (ntohs(vlan->h_vlan_TCI) &
1149 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1151 key_vlan->vlan_tpid = saved_vlan_tpid;
1152 key_vlan->vlan_eth_type = proto;
1155 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1158 case htons(ETH_P_PPP_SES): {
1160 struct pppoe_hdr hdr;
1163 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1165 fdret = FLOW_DISSECT_RET_OUT_BAD;
1170 nhoff += PPPOE_SES_HLEN;
1173 proto = htons(ETH_P_IP);
1174 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1176 case htons(PPP_IPV6):
1177 proto = htons(ETH_P_IPV6);
1178 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1181 fdret = FLOW_DISSECT_RET_OUT_BAD;
1186 case htons(ETH_P_TIPC): {
1187 struct tipc_basic_hdr *hdr, _hdr;
1189 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1192 fdret = FLOW_DISSECT_RET_OUT_BAD;
1196 if (dissector_uses_key(flow_dissector,
1197 FLOW_DISSECTOR_KEY_TIPC)) {
1198 key_addrs = skb_flow_dissector_target(flow_dissector,
1199 FLOW_DISSECTOR_KEY_TIPC,
1201 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1202 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1204 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1208 case htons(ETH_P_MPLS_UC):
1209 case htons(ETH_P_MPLS_MC):
1210 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1211 target_container, data,
1214 case htons(ETH_P_FCOE):
1215 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1216 fdret = FLOW_DISSECT_RET_OUT_BAD;
1220 nhoff += FCOE_HEADER_LEN;
1221 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1224 case htons(ETH_P_ARP):
1225 case htons(ETH_P_RARP):
1226 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1227 target_container, data,
1231 case htons(ETH_P_BATMAN):
1232 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1233 &proto, &nhoff, hlen, flags);
1237 fdret = FLOW_DISSECT_RET_OUT_BAD;
1241 /* Process result of proto processing */
1243 case FLOW_DISSECT_RET_OUT_GOOD:
1245 case FLOW_DISSECT_RET_PROTO_AGAIN:
1246 if (skb_flow_dissect_allowed(&num_hdrs))
1249 case FLOW_DISSECT_RET_CONTINUE:
1250 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1252 case FLOW_DISSECT_RET_OUT_BAD:
1258 fdret = FLOW_DISSECT_RET_CONTINUE;
1262 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1263 target_container, data,
1264 &proto, &nhoff, &hlen, flags);
1268 case NEXTHDR_ROUTING:
1269 case NEXTHDR_DEST: {
1270 u8 _opthdr[2], *opthdr;
1272 if (proto != htons(ETH_P_IPV6))
1275 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1276 data, hlen, &_opthdr);
1278 fdret = FLOW_DISSECT_RET_OUT_BAD;
1282 ip_proto = opthdr[0];
1283 nhoff += (opthdr[1] + 1) << 3;
1285 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1288 case NEXTHDR_FRAGMENT: {
1289 struct frag_hdr _fh, *fh;
1291 if (proto != htons(ETH_P_IPV6))
1294 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1298 fdret = FLOW_DISSECT_RET_OUT_BAD;
1302 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1304 nhoff += sizeof(_fh);
1305 ip_proto = fh->nexthdr;
1307 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1308 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1309 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1310 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1315 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1319 proto = htons(ETH_P_IP);
1321 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1322 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1323 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1327 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1331 proto = htons(ETH_P_IPV6);
1333 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1334 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1335 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1339 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1344 proto = htons(ETH_P_MPLS_UC);
1345 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1349 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1357 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1358 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1359 data, nhoff, ip_proto, hlen);
1361 if (dissector_uses_key(flow_dissector,
1362 FLOW_DISSECTOR_KEY_ICMP)) {
1363 key_icmp = skb_flow_dissector_target(flow_dissector,
1364 FLOW_DISSECTOR_KEY_ICMP,
1366 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1369 /* Process result of IP proto processing */
1371 case FLOW_DISSECT_RET_PROTO_AGAIN:
1372 if (skb_flow_dissect_allowed(&num_hdrs))
1375 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1376 if (skb_flow_dissect_allowed(&num_hdrs))
1377 goto ip_proto_again;
1379 case FLOW_DISSECT_RET_OUT_GOOD:
1380 case FLOW_DISSECT_RET_CONTINUE:
1382 case FLOW_DISSECT_RET_OUT_BAD:
1391 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1392 key_basic->n_proto = proto;
1393 key_basic->ip_proto = ip_proto;
1401 EXPORT_SYMBOL(__skb_flow_dissect);
1403 static siphash_key_t hashrnd __read_mostly;
1404 static __always_inline void __flow_hash_secret_init(void)
1406 net_get_random_once(&hashrnd, sizeof(hashrnd));
1409 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1411 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1412 return &flow->FLOW_KEYS_HASH_START_FIELD;
1415 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1417 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1418 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1419 sizeof(*flow) - sizeof(flow->addrs));
1421 switch (flow->control.addr_type) {
1422 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1423 diff -= sizeof(flow->addrs.v4addrs);
1425 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1426 diff -= sizeof(flow->addrs.v6addrs);
1428 case FLOW_DISSECTOR_KEY_TIPC:
1429 diff -= sizeof(flow->addrs.tipckey);
1432 return sizeof(*flow) - diff;
1435 __be32 flow_get_u32_src(const struct flow_keys *flow)
1437 switch (flow->control.addr_type) {
1438 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1439 return flow->addrs.v4addrs.src;
1440 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1441 return (__force __be32)ipv6_addr_hash(
1442 &flow->addrs.v6addrs.src);
1443 case FLOW_DISSECTOR_KEY_TIPC:
1444 return flow->addrs.tipckey.key;
1449 EXPORT_SYMBOL(flow_get_u32_src);
1451 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1453 switch (flow->control.addr_type) {
1454 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1455 return flow->addrs.v4addrs.dst;
1456 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1457 return (__force __be32)ipv6_addr_hash(
1458 &flow->addrs.v6addrs.dst);
1463 EXPORT_SYMBOL(flow_get_u32_dst);
1465 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1469 switch (keys->control.addr_type) {
1470 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1471 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1472 (__force u32)keys->addrs.v4addrs.src;
1473 if ((addr_diff < 0) ||
1475 ((__force u16)keys->ports.dst <
1476 (__force u16)keys->ports.src))) {
1477 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1478 swap(keys->ports.src, keys->ports.dst);
1481 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1482 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1483 &keys->addrs.v6addrs.src,
1484 sizeof(keys->addrs.v6addrs.dst));
1485 if ((addr_diff < 0) ||
1487 ((__force u16)keys->ports.dst <
1488 (__force u16)keys->ports.src))) {
1489 for (i = 0; i < 4; i++)
1490 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1491 keys->addrs.v6addrs.dst.s6_addr32[i]);
1492 swap(keys->ports.src, keys->ports.dst);
1498 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1499 const siphash_key_t *keyval)
1503 __flow_hash_consistentify(keys);
1505 hash = siphash(flow_keys_hash_start(keys),
1506 flow_keys_hash_length(keys), keyval);
1513 u32 flow_hash_from_keys(struct flow_keys *keys)
1515 __flow_hash_secret_init();
1516 return __flow_hash_from_keys(keys, &hashrnd);
1518 EXPORT_SYMBOL(flow_hash_from_keys);
1520 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1521 struct flow_keys *keys,
1522 const siphash_key_t *keyval)
1524 skb_flow_dissect_flow_keys(skb, keys,
1525 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1527 return __flow_hash_from_keys(keys, keyval);
1530 struct _flow_keys_digest_data {
1539 void make_flow_keys_digest(struct flow_keys_digest *digest,
1540 const struct flow_keys *flow)
1542 struct _flow_keys_digest_data *data =
1543 (struct _flow_keys_digest_data *)digest;
1545 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1547 memset(digest, 0, sizeof(*digest));
1549 data->n_proto = flow->basic.n_proto;
1550 data->ip_proto = flow->basic.ip_proto;
1551 data->ports = flow->ports.ports;
1552 data->src = flow->addrs.v4addrs.src;
1553 data->dst = flow->addrs.v4addrs.dst;
1555 EXPORT_SYMBOL(make_flow_keys_digest);
1557 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1559 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1561 struct flow_keys keys;
1563 __flow_hash_secret_init();
1565 memset(&keys, 0, sizeof(keys));
1566 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1567 &keys, NULL, 0, 0, 0, 0);
1569 return __flow_hash_from_keys(&keys, &hashrnd);
1571 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1574 * __skb_get_hash: calculate a flow hash
1575 * @skb: sk_buff to calculate flow hash from
1577 * This function calculates a flow hash based on src/dst addresses
1578 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1579 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1580 * if hash is a canonical 4-tuple hash over transport ports.
1582 void __skb_get_hash(struct sk_buff *skb)
1584 struct flow_keys keys;
1587 __flow_hash_secret_init();
1589 hash = ___skb_get_hash(skb, &keys, &hashrnd);
1591 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1593 EXPORT_SYMBOL(__skb_get_hash);
1595 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1596 const siphash_key_t *perturb)
1598 struct flow_keys keys;
1600 return ___skb_get_hash(skb, &keys, perturb);
1602 EXPORT_SYMBOL(skb_get_hash_perturb);
1604 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1605 const struct flow_keys_basic *keys, int hlen)
1607 u32 poff = keys->control.thoff;
1609 /* skip L4 headers for fragments after the first */
1610 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1611 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1614 switch (keys->basic.ip_proto) {
1616 /* access doff as u8 to avoid unaligned access */
1620 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1621 data, hlen, &_doff);
1625 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1629 case IPPROTO_UDPLITE:
1630 poff += sizeof(struct udphdr);
1632 /* For the rest, we do not really care about header
1633 * extensions at this point for now.
1636 poff += sizeof(struct icmphdr);
1638 case IPPROTO_ICMPV6:
1639 poff += sizeof(struct icmp6hdr);
1642 poff += sizeof(struct igmphdr);
1645 poff += sizeof(struct dccp_hdr);
1648 poff += sizeof(struct sctphdr);
1656 * skb_get_poff - get the offset to the payload
1657 * @skb: sk_buff to get the payload offset from
1659 * The function will get the offset to the payload as far as it could
1660 * be dissected. The main user is currently BPF, so that we can dynamically
1661 * truncate packets without needing to push actual payload to the user
1662 * space and can analyze headers only, instead.
1664 u32 skb_get_poff(const struct sk_buff *skb)
1666 struct flow_keys_basic keys;
1668 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1672 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1675 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1677 memset(keys, 0, sizeof(*keys));
1679 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1680 sizeof(keys->addrs.v6addrs.src));
1681 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1682 sizeof(keys->addrs.v6addrs.dst));
1683 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1684 keys->ports.src = fl6->fl6_sport;
1685 keys->ports.dst = fl6->fl6_dport;
1686 keys->keyid.keyid = fl6->fl6_gre_key;
1687 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1688 keys->basic.ip_proto = fl6->flowi6_proto;
1690 return flow_hash_from_keys(keys);
1692 EXPORT_SYMBOL(__get_hash_from_flowi6);
1694 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1696 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1697 .offset = offsetof(struct flow_keys, control),
1700 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1701 .offset = offsetof(struct flow_keys, basic),
1704 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1705 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1708 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1709 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1712 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1713 .offset = offsetof(struct flow_keys, addrs.tipckey),
1716 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1717 .offset = offsetof(struct flow_keys, ports),
1720 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1721 .offset = offsetof(struct flow_keys, vlan),
1724 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1725 .offset = offsetof(struct flow_keys, tags),
1728 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1729 .offset = offsetof(struct flow_keys, keyid),
1733 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1735 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1736 .offset = offsetof(struct flow_keys, control),
1739 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1740 .offset = offsetof(struct flow_keys, basic),
1743 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1744 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1747 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1748 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1751 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1752 .offset = offsetof(struct flow_keys, ports),
1756 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1758 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1759 .offset = offsetof(struct flow_keys, control),
1762 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1763 .offset = offsetof(struct flow_keys, basic),
1767 struct flow_dissector flow_keys_dissector __read_mostly;
1768 EXPORT_SYMBOL(flow_keys_dissector);
1770 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1771 EXPORT_SYMBOL(flow_keys_basic_dissector);
1773 static int __init init_default_flow_dissectors(void)
1775 skb_flow_dissector_init(&flow_keys_dissector,
1776 flow_keys_dissector_keys,
1777 ARRAY_SIZE(flow_keys_dissector_keys));
1778 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1779 flow_keys_dissector_symmetric_keys,
1780 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1781 skb_flow_dissector_init(&flow_keys_basic_dissector,
1782 flow_keys_basic_dissector_keys,
1783 ARRAY_SIZE(flow_keys_basic_dissector_keys));
1785 return register_pernet_subsys(&flow_dissector_pernet_ops);
1787 core_initcall(init_default_flow_dissectors);