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>
8 #include <linux/filter.h>
10 #include <net/dst_metadata.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 <scsi/fc/fc_fcoe.h>
31 #include <uapi/linux/batadv_packet.h>
32 #include <linux/bpf.h>
33 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
34 #include <net/netfilter/nf_conntrack_core.h>
35 #include <net/netfilter/nf_conntrack_labels.h>
37 #include <linux/bpf-netns.h>
39 static void dissector_set_key(struct flow_dissector *flow_dissector,
40 enum flow_dissector_key_id key_id)
42 flow_dissector->used_keys |= (1 << key_id);
45 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
46 const struct flow_dissector_key *key,
47 unsigned int key_count)
51 memset(flow_dissector, 0, sizeof(*flow_dissector));
53 for (i = 0; i < key_count; i++, key++) {
54 /* User should make sure that every key target offset is within
55 * boundaries of unsigned short.
57 BUG_ON(key->offset > USHRT_MAX);
58 BUG_ON(dissector_uses_key(flow_dissector,
61 dissector_set_key(flow_dissector, key->key_id);
62 flow_dissector->offset[key->key_id] = key->offset;
65 /* Ensure that the dissector always includes control and basic key.
66 * That way we are able to avoid handling lack of these in fast path.
68 BUG_ON(!dissector_uses_key(flow_dissector,
69 FLOW_DISSECTOR_KEY_CONTROL));
70 BUG_ON(!dissector_uses_key(flow_dissector,
71 FLOW_DISSECTOR_KEY_BASIC));
73 EXPORT_SYMBOL(skb_flow_dissector_init);
75 #ifdef CONFIG_BPF_SYSCALL
76 int flow_dissector_bpf_prog_attach_check(struct net *net,
77 struct bpf_prog *prog)
79 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
81 if (net == &init_net) {
82 /* BPF flow dissector in the root namespace overrides
83 * any per-net-namespace one. When attaching to root,
84 * make sure we don't have any BPF program attached
85 * to the non-root namespaces.
92 if (rcu_access_pointer(ns->bpf.run_array[type]))
96 /* Make sure root flow dissector is not attached
97 * when attaching to the non-root namespace.
99 if (rcu_access_pointer(init_net.bpf.run_array[type]))
105 #endif /* CONFIG_BPF_SYSCALL */
108 * __skb_flow_get_ports - extract the upper layer ports and return them
109 * @skb: sk_buff to extract the ports from
110 * @thoff: transport header offset
111 * @ip_proto: protocol for which to get port offset
112 * @data: raw buffer pointer to the packet, if NULL use skb->data
113 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
115 * The function will try to retrieve the ports at offset thoff + poff where poff
116 * is the protocol port offset returned from proto_ports_offset
118 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
119 const void *data, int hlen)
121 int poff = proto_ports_offset(ip_proto);
125 hlen = skb_headlen(skb);
129 __be32 *ports, _ports;
131 ports = __skb_header_pointer(skb, thoff + poff,
132 sizeof(_ports), data, hlen, &_ports);
139 EXPORT_SYMBOL(__skb_flow_get_ports);
141 static bool icmp_has_id(u8 type)
147 case ICMP_TIMESTAMPREPLY:
148 case ICMPV6_ECHO_REQUEST:
149 case ICMPV6_ECHO_REPLY:
157 * skb_flow_get_icmp_tci - extract ICMP(6) Type, Code and Identifier fields
158 * @skb: sk_buff to extract from
159 * @key_icmp: struct flow_dissector_key_icmp to fill
160 * @data: raw buffer pointer to the packet
161 * @thoff: offset to extract at
162 * @hlen: packet header length
164 void skb_flow_get_icmp_tci(const struct sk_buff *skb,
165 struct flow_dissector_key_icmp *key_icmp,
166 const void *data, int thoff, int hlen)
168 struct icmphdr *ih, _ih;
170 ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
174 key_icmp->type = ih->type;
175 key_icmp->code = ih->code;
177 /* As we use 0 to signal that the Id field is not present,
178 * avoid confusion with packets without such field
180 if (icmp_has_id(ih->type))
181 key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
185 EXPORT_SYMBOL(skb_flow_get_icmp_tci);
187 /* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
188 * using skb_flow_get_icmp_tci().
190 static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
191 struct flow_dissector *flow_dissector,
192 void *target_container, const void *data,
195 struct flow_dissector_key_icmp *key_icmp;
197 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
200 key_icmp = skb_flow_dissector_target(flow_dissector,
201 FLOW_DISSECTOR_KEY_ICMP,
204 skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
207 void skb_flow_dissect_meta(const struct sk_buff *skb,
208 struct flow_dissector *flow_dissector,
209 void *target_container)
211 struct flow_dissector_key_meta *meta;
213 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
216 meta = skb_flow_dissector_target(flow_dissector,
217 FLOW_DISSECTOR_KEY_META,
219 meta->ingress_ifindex = skb->skb_iif;
221 EXPORT_SYMBOL(skb_flow_dissect_meta);
224 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
225 struct flow_dissector *flow_dissector,
226 void *target_container)
228 struct flow_dissector_key_control *ctrl;
230 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
233 ctrl = skb_flow_dissector_target(flow_dissector,
234 FLOW_DISSECTOR_KEY_ENC_CONTROL,
236 ctrl->addr_type = type;
240 skb_flow_dissect_ct(const struct sk_buff *skb,
241 struct flow_dissector *flow_dissector,
242 void *target_container, u16 *ctinfo_map,
243 size_t mapsize, bool post_ct, u16 zone)
245 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
246 struct flow_dissector_key_ct *key;
247 enum ip_conntrack_info ctinfo;
248 struct nf_conn_labels *cl;
251 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
254 ct = nf_ct_get(skb, &ctinfo);
258 key = skb_flow_dissector_target(flow_dissector,
259 FLOW_DISSECTOR_KEY_CT,
263 key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
264 TCA_FLOWER_KEY_CT_FLAGS_INVALID;
269 if (ctinfo < mapsize)
270 key->ct_state = ctinfo_map[ctinfo];
271 #if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
272 key->ct_zone = ct->zone.id;
274 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
275 key->ct_mark = ct->mark;
278 cl = nf_ct_labels_find(ct);
280 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
281 #endif /* CONFIG_NF_CONNTRACK */
283 EXPORT_SYMBOL(skb_flow_dissect_ct);
286 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
287 struct flow_dissector *flow_dissector,
288 void *target_container)
290 struct ip_tunnel_info *info;
291 struct ip_tunnel_key *key;
293 /* A quick check to see if there might be something to do. */
294 if (!dissector_uses_key(flow_dissector,
295 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
296 !dissector_uses_key(flow_dissector,
297 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
298 !dissector_uses_key(flow_dissector,
299 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
300 !dissector_uses_key(flow_dissector,
301 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
302 !dissector_uses_key(flow_dissector,
303 FLOW_DISSECTOR_KEY_ENC_PORTS) &&
304 !dissector_uses_key(flow_dissector,
305 FLOW_DISSECTOR_KEY_ENC_IP) &&
306 !dissector_uses_key(flow_dissector,
307 FLOW_DISSECTOR_KEY_ENC_OPTS))
310 info = skb_tunnel_info(skb);
316 switch (ip_tunnel_info_af(info)) {
318 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
321 if (dissector_uses_key(flow_dissector,
322 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
323 struct flow_dissector_key_ipv4_addrs *ipv4;
325 ipv4 = skb_flow_dissector_target(flow_dissector,
326 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
328 ipv4->src = key->u.ipv4.src;
329 ipv4->dst = key->u.ipv4.dst;
333 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
336 if (dissector_uses_key(flow_dissector,
337 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
338 struct flow_dissector_key_ipv6_addrs *ipv6;
340 ipv6 = skb_flow_dissector_target(flow_dissector,
341 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
343 ipv6->src = key->u.ipv6.src;
344 ipv6->dst = key->u.ipv6.dst;
349 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
350 struct flow_dissector_key_keyid *keyid;
352 keyid = skb_flow_dissector_target(flow_dissector,
353 FLOW_DISSECTOR_KEY_ENC_KEYID,
355 keyid->keyid = tunnel_id_to_key32(key->tun_id);
358 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
359 struct flow_dissector_key_ports *tp;
361 tp = skb_flow_dissector_target(flow_dissector,
362 FLOW_DISSECTOR_KEY_ENC_PORTS,
364 tp->src = key->tp_src;
365 tp->dst = key->tp_dst;
368 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
369 struct flow_dissector_key_ip *ip;
371 ip = skb_flow_dissector_target(flow_dissector,
372 FLOW_DISSECTOR_KEY_ENC_IP,
378 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
379 struct flow_dissector_key_enc_opts *enc_opt;
381 enc_opt = skb_flow_dissector_target(flow_dissector,
382 FLOW_DISSECTOR_KEY_ENC_OPTS,
385 if (info->options_len) {
386 enc_opt->len = info->options_len;
387 ip_tunnel_info_opts_get(enc_opt->data, info);
388 enc_opt->dst_opt_type = info->key.tun_flags &
389 TUNNEL_OPTIONS_PRESENT;
393 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
395 void skb_flow_dissect_hash(const struct sk_buff *skb,
396 struct flow_dissector *flow_dissector,
397 void *target_container)
399 struct flow_dissector_key_hash *key;
401 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
404 key = skb_flow_dissector_target(flow_dissector,
405 FLOW_DISSECTOR_KEY_HASH,
408 key->hash = skb_get_hash_raw(skb);
410 EXPORT_SYMBOL(skb_flow_dissect_hash);
412 static enum flow_dissect_ret
413 __skb_flow_dissect_mpls(const struct sk_buff *skb,
414 struct flow_dissector *flow_dissector,
415 void *target_container, const void *data, int nhoff,
416 int hlen, int lse_index, bool *entropy_label)
418 struct mpls_label *hdr, _hdr;
419 u32 entry, label, bos;
421 if (!dissector_uses_key(flow_dissector,
422 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
423 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
424 return FLOW_DISSECT_RET_OUT_GOOD;
426 if (lse_index >= FLOW_DIS_MPLS_MAX)
427 return FLOW_DISSECT_RET_OUT_GOOD;
429 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
432 return FLOW_DISSECT_RET_OUT_BAD;
434 entry = ntohl(hdr->entry);
435 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
436 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
438 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
439 struct flow_dissector_key_mpls *key_mpls;
440 struct flow_dissector_mpls_lse *lse;
442 key_mpls = skb_flow_dissector_target(flow_dissector,
443 FLOW_DISSECTOR_KEY_MPLS,
445 lse = &key_mpls->ls[lse_index];
447 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
449 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
450 lse->mpls_label = label;
451 dissector_set_mpls_lse(key_mpls, lse_index);
454 if (*entropy_label &&
455 dissector_uses_key(flow_dissector,
456 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
457 struct flow_dissector_key_keyid *key_keyid;
459 key_keyid = skb_flow_dissector_target(flow_dissector,
460 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
462 key_keyid->keyid = cpu_to_be32(label);
465 *entropy_label = label == MPLS_LABEL_ENTROPY;
467 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
470 static enum flow_dissect_ret
471 __skb_flow_dissect_arp(const struct sk_buff *skb,
472 struct flow_dissector *flow_dissector,
473 void *target_container, const void *data,
476 struct flow_dissector_key_arp *key_arp;
478 unsigned char ar_sha[ETH_ALEN];
479 unsigned char ar_sip[4];
480 unsigned char ar_tha[ETH_ALEN];
481 unsigned char ar_tip[4];
482 } *arp_eth, _arp_eth;
483 const struct arphdr *arp;
486 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
487 return FLOW_DISSECT_RET_OUT_GOOD;
489 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
492 return FLOW_DISSECT_RET_OUT_BAD;
494 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
495 arp->ar_pro != htons(ETH_P_IP) ||
496 arp->ar_hln != ETH_ALEN ||
498 (arp->ar_op != htons(ARPOP_REPLY) &&
499 arp->ar_op != htons(ARPOP_REQUEST)))
500 return FLOW_DISSECT_RET_OUT_BAD;
502 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
503 sizeof(_arp_eth), data,
506 return FLOW_DISSECT_RET_OUT_BAD;
508 key_arp = skb_flow_dissector_target(flow_dissector,
509 FLOW_DISSECTOR_KEY_ARP,
512 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
513 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
515 /* Only store the lower byte of the opcode;
516 * this covers ARPOP_REPLY and ARPOP_REQUEST.
518 key_arp->op = ntohs(arp->ar_op) & 0xff;
520 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
521 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
523 return FLOW_DISSECT_RET_OUT_GOOD;
526 static enum flow_dissect_ret
527 __skb_flow_dissect_gre(const struct sk_buff *skb,
528 struct flow_dissector_key_control *key_control,
529 struct flow_dissector *flow_dissector,
530 void *target_container, const void *data,
531 __be16 *p_proto, int *p_nhoff, int *p_hlen,
534 struct flow_dissector_key_keyid *key_keyid;
535 struct gre_base_hdr *hdr, _hdr;
539 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
540 data, *p_hlen, &_hdr);
542 return FLOW_DISSECT_RET_OUT_BAD;
544 /* Only look inside GRE without routing */
545 if (hdr->flags & GRE_ROUTING)
546 return FLOW_DISSECT_RET_OUT_GOOD;
548 /* Only look inside GRE for version 0 and 1 */
549 gre_ver = ntohs(hdr->flags & GRE_VERSION);
551 return FLOW_DISSECT_RET_OUT_GOOD;
553 *p_proto = hdr->protocol;
555 /* Version1 must be PPTP, and check the flags */
556 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
557 return FLOW_DISSECT_RET_OUT_GOOD;
560 offset += sizeof(struct gre_base_hdr);
562 if (hdr->flags & GRE_CSUM)
563 offset += sizeof_field(struct gre_full_hdr, csum) +
564 sizeof_field(struct gre_full_hdr, reserved1);
566 if (hdr->flags & GRE_KEY) {
570 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
572 data, *p_hlen, &_keyid);
574 return FLOW_DISSECT_RET_OUT_BAD;
576 if (dissector_uses_key(flow_dissector,
577 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
578 key_keyid = skb_flow_dissector_target(flow_dissector,
579 FLOW_DISSECTOR_KEY_GRE_KEYID,
582 key_keyid->keyid = *keyid;
584 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
586 offset += sizeof_field(struct gre_full_hdr, key);
589 if (hdr->flags & GRE_SEQ)
590 offset += sizeof_field(struct pptp_gre_header, seq);
593 if (*p_proto == htons(ETH_P_TEB)) {
594 const struct ethhdr *eth;
597 eth = __skb_header_pointer(skb, *p_nhoff + offset,
599 data, *p_hlen, &_eth);
601 return FLOW_DISSECT_RET_OUT_BAD;
602 *p_proto = eth->h_proto;
603 offset += sizeof(*eth);
605 /* Cap headers that we access via pointers at the
606 * end of the Ethernet header as our maximum alignment
607 * at that point is only 2 bytes.
610 *p_hlen = *p_nhoff + offset;
612 } else { /* version 1, must be PPTP */
613 u8 _ppp_hdr[PPP_HDRLEN];
616 if (hdr->flags & GRE_ACK)
617 offset += sizeof_field(struct pptp_gre_header, ack);
619 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
621 data, *p_hlen, _ppp_hdr);
623 return FLOW_DISSECT_RET_OUT_BAD;
625 switch (PPP_PROTOCOL(ppp_hdr)) {
627 *p_proto = htons(ETH_P_IP);
630 *p_proto = htons(ETH_P_IPV6);
633 /* Could probably catch some more like MPLS */
637 offset += PPP_HDRLEN;
641 key_control->flags |= FLOW_DIS_ENCAPSULATION;
642 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
643 return FLOW_DISSECT_RET_OUT_GOOD;
645 return FLOW_DISSECT_RET_PROTO_AGAIN;
649 * __skb_flow_dissect_batadv() - dissect batman-adv header
650 * @skb: sk_buff to with the batman-adv header
651 * @key_control: flow dissectors control key
652 * @data: raw buffer pointer to the packet, if NULL use skb->data
653 * @p_proto: pointer used to update the protocol to process next
654 * @p_nhoff: pointer used to update inner network header offset
655 * @hlen: packet header length
656 * @flags: any combination of FLOW_DISSECTOR_F_*
658 * ETH_P_BATMAN packets are tried to be dissected. Only
659 * &struct batadv_unicast packets are actually processed because they contain an
660 * inner ethernet header and are usually followed by actual network header. This
661 * allows the flow dissector to continue processing the packet.
663 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
664 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
665 * otherwise FLOW_DISSECT_RET_OUT_BAD
667 static enum flow_dissect_ret
668 __skb_flow_dissect_batadv(const struct sk_buff *skb,
669 struct flow_dissector_key_control *key_control,
670 const void *data, __be16 *p_proto, int *p_nhoff,
671 int hlen, unsigned int flags)
674 struct batadv_unicast_packet batadv_unicast;
678 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
681 return FLOW_DISSECT_RET_OUT_BAD;
683 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
684 return FLOW_DISSECT_RET_OUT_BAD;
686 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
687 return FLOW_DISSECT_RET_OUT_BAD;
689 *p_proto = hdr->eth.h_proto;
690 *p_nhoff += sizeof(*hdr);
692 key_control->flags |= FLOW_DIS_ENCAPSULATION;
693 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
694 return FLOW_DISSECT_RET_OUT_GOOD;
696 return FLOW_DISSECT_RET_PROTO_AGAIN;
700 __skb_flow_dissect_tcp(const struct sk_buff *skb,
701 struct flow_dissector *flow_dissector,
702 void *target_container, const void *data,
705 struct flow_dissector_key_tcp *key_tcp;
706 struct tcphdr *th, _th;
708 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
711 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
715 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
718 key_tcp = skb_flow_dissector_target(flow_dissector,
719 FLOW_DISSECTOR_KEY_TCP,
721 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
725 __skb_flow_dissect_ports(const struct sk_buff *skb,
726 struct flow_dissector *flow_dissector,
727 void *target_container, const void *data,
728 int nhoff, u8 ip_proto, int hlen)
730 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
731 struct flow_dissector_key_ports *key_ports;
733 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
734 dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
735 else if (dissector_uses_key(flow_dissector,
736 FLOW_DISSECTOR_KEY_PORTS_RANGE))
737 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
739 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
742 key_ports = skb_flow_dissector_target(flow_dissector,
745 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
750 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
751 struct flow_dissector *flow_dissector,
752 void *target_container, const void *data,
753 const struct iphdr *iph)
755 struct flow_dissector_key_ip *key_ip;
757 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
760 key_ip = skb_flow_dissector_target(flow_dissector,
761 FLOW_DISSECTOR_KEY_IP,
763 key_ip->tos = iph->tos;
764 key_ip->ttl = iph->ttl;
768 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
769 struct flow_dissector *flow_dissector,
770 void *target_container, const void *data,
771 const struct ipv6hdr *iph)
773 struct flow_dissector_key_ip *key_ip;
775 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
778 key_ip = skb_flow_dissector_target(flow_dissector,
779 FLOW_DISSECTOR_KEY_IP,
781 key_ip->tos = ipv6_get_dsfield(iph);
782 key_ip->ttl = iph->hop_limit;
785 /* Maximum number of protocol headers that can be parsed in
788 #define MAX_FLOW_DISSECT_HDRS 15
790 static bool skb_flow_dissect_allowed(int *num_hdrs)
794 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
797 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
798 struct flow_dissector *flow_dissector,
799 void *target_container)
801 struct flow_dissector_key_ports *key_ports = NULL;
802 struct flow_dissector_key_control *key_control;
803 struct flow_dissector_key_basic *key_basic;
804 struct flow_dissector_key_addrs *key_addrs;
805 struct flow_dissector_key_tags *key_tags;
807 key_control = skb_flow_dissector_target(flow_dissector,
808 FLOW_DISSECTOR_KEY_CONTROL,
810 key_control->thoff = flow_keys->thoff;
811 if (flow_keys->is_frag)
812 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
813 if (flow_keys->is_first_frag)
814 key_control->flags |= FLOW_DIS_FIRST_FRAG;
815 if (flow_keys->is_encap)
816 key_control->flags |= FLOW_DIS_ENCAPSULATION;
818 key_basic = skb_flow_dissector_target(flow_dissector,
819 FLOW_DISSECTOR_KEY_BASIC,
821 key_basic->n_proto = flow_keys->n_proto;
822 key_basic->ip_proto = flow_keys->ip_proto;
824 if (flow_keys->addr_proto == ETH_P_IP &&
825 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
826 key_addrs = skb_flow_dissector_target(flow_dissector,
827 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
829 key_addrs->v4addrs.src = flow_keys->ipv4_src;
830 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
831 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
832 } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
833 dissector_uses_key(flow_dissector,
834 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
835 key_addrs = skb_flow_dissector_target(flow_dissector,
836 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
838 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
839 sizeof(key_addrs->v6addrs.src));
840 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
841 sizeof(key_addrs->v6addrs.dst));
842 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
845 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
846 key_ports = skb_flow_dissector_target(flow_dissector,
847 FLOW_DISSECTOR_KEY_PORTS,
849 else if (dissector_uses_key(flow_dissector,
850 FLOW_DISSECTOR_KEY_PORTS_RANGE))
851 key_ports = skb_flow_dissector_target(flow_dissector,
852 FLOW_DISSECTOR_KEY_PORTS_RANGE,
856 key_ports->src = flow_keys->sport;
857 key_ports->dst = flow_keys->dport;
860 if (dissector_uses_key(flow_dissector,
861 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
862 key_tags = skb_flow_dissector_target(flow_dissector,
863 FLOW_DISSECTOR_KEY_FLOW_LABEL,
865 key_tags->flow_label = ntohl(flow_keys->flow_label);
869 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
870 __be16 proto, int nhoff, int hlen, unsigned int flags)
872 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
875 /* Pass parameters to the BPF program */
876 memset(flow_keys, 0, sizeof(*flow_keys));
877 flow_keys->n_proto = proto;
878 flow_keys->nhoff = nhoff;
879 flow_keys->thoff = flow_keys->nhoff;
881 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
882 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
883 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
884 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
885 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
886 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
887 flow_keys->flags = flags;
889 result = bpf_prog_run_pin_on_cpu(prog, ctx);
891 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
892 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
893 flow_keys->nhoff, hlen);
895 return result == BPF_OK;
899 * __skb_flow_dissect - extract the flow_keys struct and return it
900 * @net: associated network namespace, derived from @skb if NULL
901 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
902 * @flow_dissector: list of keys to dissect
903 * @target_container: target structure to put dissected values into
904 * @data: raw buffer pointer to the packet, if NULL use skb->data
905 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
906 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
907 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
908 * @flags: flags that control the dissection process, e.g.
909 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
911 * The function will try to retrieve individual keys into target specified
912 * by flow_dissector from either the skbuff or a raw buffer specified by the
915 * Caller must take care of zeroing target container memory.
917 bool __skb_flow_dissect(const struct net *net,
918 const struct sk_buff *skb,
919 struct flow_dissector *flow_dissector,
920 void *target_container, const void *data,
921 __be16 proto, int nhoff, int hlen, unsigned int flags)
923 struct flow_dissector_key_control *key_control;
924 struct flow_dissector_key_basic *key_basic;
925 struct flow_dissector_key_addrs *key_addrs;
926 struct flow_dissector_key_tags *key_tags;
927 struct flow_dissector_key_vlan *key_vlan;
928 enum flow_dissect_ret fdret;
929 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
930 bool mpls_el = false;
938 proto = skb_vlan_tag_present(skb) ?
939 skb->vlan_proto : skb->protocol;
940 nhoff = skb_network_offset(skb);
941 hlen = skb_headlen(skb);
942 #if IS_ENABLED(CONFIG_NET_DSA)
943 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
944 proto == htons(ETH_P_XDSA))) {
945 const struct dsa_device_ops *ops;
948 ops = skb->dev->dsa_ptr->tag_ops;
949 /* Only DSA header taggers break flow dissection */
950 if (ops->needed_headroom) {
951 if (ops->flow_dissect)
952 ops->flow_dissect(skb, &proto, &offset);
954 dsa_tag_generic_flow_dissect(skb,
964 /* It is ensured by skb_flow_dissector_init() that control key will
967 key_control = skb_flow_dissector_target(flow_dissector,
968 FLOW_DISSECTOR_KEY_CONTROL,
971 /* It is ensured by skb_flow_dissector_init() that basic key will
974 key_basic = skb_flow_dissector_target(flow_dissector,
975 FLOW_DISSECTOR_KEY_BASIC,
981 net = dev_net(skb->dev);
983 net = sock_net(skb->sk);
989 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
990 struct bpf_prog_array *run_array;
993 run_array = rcu_dereference(init_net.bpf.run_array[type]);
995 run_array = rcu_dereference(net->bpf.run_array[type]);
998 struct bpf_flow_keys flow_keys;
999 struct bpf_flow_dissector ctx = {
1000 .flow_keys = &flow_keys,
1002 .data_end = data + hlen,
1004 __be16 n_proto = proto;
1005 struct bpf_prog *prog;
1009 /* we can't use 'proto' in the skb case
1010 * because it might be set to skb->vlan_proto
1011 * which has been pulled from the data
1013 n_proto = skb->protocol;
1016 prog = READ_ONCE(run_array->items[0].prog);
1017 ret = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1019 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1027 if (dissector_uses_key(flow_dissector,
1028 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1029 struct ethhdr *eth = eth_hdr(skb);
1030 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1032 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1033 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1035 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1038 if (dissector_uses_key(flow_dissector,
1039 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1040 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1042 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1043 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1045 key_num_of_vlans->num_of_vlans = 0;
1049 fdret = FLOW_DISSECT_RET_CONTINUE;
1052 case htons(ETH_P_IP): {
1053 const struct iphdr *iph;
1056 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1057 if (!iph || iph->ihl < 5) {
1058 fdret = FLOW_DISSECT_RET_OUT_BAD;
1062 nhoff += iph->ihl * 4;
1064 ip_proto = iph->protocol;
1066 if (dissector_uses_key(flow_dissector,
1067 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1068 key_addrs = skb_flow_dissector_target(flow_dissector,
1069 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1072 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1073 sizeof(key_addrs->v4addrs.src));
1074 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1075 sizeof(key_addrs->v4addrs.dst));
1076 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1079 __skb_flow_dissect_ipv4(skb, flow_dissector,
1080 target_container, data, iph);
1082 if (ip_is_fragment(iph)) {
1083 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1085 if (iph->frag_off & htons(IP_OFFSET)) {
1086 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1089 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1091 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1092 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1100 case htons(ETH_P_IPV6): {
1101 const struct ipv6hdr *iph;
1102 struct ipv6hdr _iph;
1104 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1106 fdret = FLOW_DISSECT_RET_OUT_BAD;
1110 ip_proto = iph->nexthdr;
1111 nhoff += sizeof(struct ipv6hdr);
1113 if (dissector_uses_key(flow_dissector,
1114 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1115 key_addrs = skb_flow_dissector_target(flow_dissector,
1116 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1119 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1120 sizeof(key_addrs->v6addrs.src));
1121 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1122 sizeof(key_addrs->v6addrs.dst));
1123 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1126 if ((dissector_uses_key(flow_dissector,
1127 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1128 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1129 ip6_flowlabel(iph)) {
1130 __be32 flow_label = ip6_flowlabel(iph);
1132 if (dissector_uses_key(flow_dissector,
1133 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1134 key_tags = skb_flow_dissector_target(flow_dissector,
1135 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1137 key_tags->flow_label = ntohl(flow_label);
1139 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1140 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1145 __skb_flow_dissect_ipv6(skb, flow_dissector,
1146 target_container, data, iph);
1150 case htons(ETH_P_8021AD):
1151 case htons(ETH_P_8021Q): {
1152 const struct vlan_hdr *vlan = NULL;
1153 struct vlan_hdr _vlan;
1154 __be16 saved_vlan_tpid = proto;
1156 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1157 skb && skb_vlan_tag_present(skb)) {
1158 proto = skb->protocol;
1160 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1161 data, hlen, &_vlan);
1163 fdret = FLOW_DISSECT_RET_OUT_BAD;
1167 proto = vlan->h_vlan_encapsulated_proto;
1168 nhoff += sizeof(*vlan);
1171 if (dissector_uses_key(flow_dissector,
1172 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1173 struct flow_dissector_key_num_of_vlans *key_nvs;
1175 key_nvs = skb_flow_dissector_target(flow_dissector,
1176 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1178 key_nvs->num_of_vlans++;
1181 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1182 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1183 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1184 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1186 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1190 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1191 key_vlan = skb_flow_dissector_target(flow_dissector,
1196 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1197 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1199 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1201 key_vlan->vlan_priority =
1202 (ntohs(vlan->h_vlan_TCI) &
1203 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1205 key_vlan->vlan_tpid = saved_vlan_tpid;
1206 key_vlan->vlan_eth_type = proto;
1209 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1212 case htons(ETH_P_PPP_SES): {
1214 struct pppoe_hdr hdr;
1217 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1219 fdret = FLOW_DISSECT_RET_OUT_BAD;
1223 nhoff += PPPOE_SES_HLEN;
1224 switch (hdr->proto) {
1226 proto = htons(ETH_P_IP);
1227 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1229 case htons(PPP_IPV6):
1230 proto = htons(ETH_P_IPV6);
1231 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1234 fdret = FLOW_DISSECT_RET_OUT_BAD;
1239 case htons(ETH_P_TIPC): {
1240 struct tipc_basic_hdr *hdr, _hdr;
1242 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1245 fdret = FLOW_DISSECT_RET_OUT_BAD;
1249 if (dissector_uses_key(flow_dissector,
1250 FLOW_DISSECTOR_KEY_TIPC)) {
1251 key_addrs = skb_flow_dissector_target(flow_dissector,
1252 FLOW_DISSECTOR_KEY_TIPC,
1254 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1255 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1257 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1261 case htons(ETH_P_MPLS_UC):
1262 case htons(ETH_P_MPLS_MC):
1263 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1264 target_container, data,
1265 nhoff, hlen, mpls_lse,
1267 nhoff += sizeof(struct mpls_label);
1270 case htons(ETH_P_FCOE):
1271 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1272 fdret = FLOW_DISSECT_RET_OUT_BAD;
1276 nhoff += FCOE_HEADER_LEN;
1277 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1280 case htons(ETH_P_ARP):
1281 case htons(ETH_P_RARP):
1282 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1283 target_container, data,
1287 case htons(ETH_P_BATMAN):
1288 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1289 &proto, &nhoff, hlen, flags);
1292 case htons(ETH_P_1588): {
1293 struct ptp_header *hdr, _hdr;
1295 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1298 fdret = FLOW_DISSECT_RET_OUT_BAD;
1302 nhoff += ntohs(hdr->message_length);
1303 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1307 case htons(ETH_P_PRP):
1308 case htons(ETH_P_HSR): {
1309 struct hsr_tag *hdr, _hdr;
1311 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1314 fdret = FLOW_DISSECT_RET_OUT_BAD;
1318 proto = hdr->encap_proto;
1320 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1325 fdret = FLOW_DISSECT_RET_OUT_BAD;
1329 /* Process result of proto processing */
1331 case FLOW_DISSECT_RET_OUT_GOOD:
1333 case FLOW_DISSECT_RET_PROTO_AGAIN:
1334 if (skb_flow_dissect_allowed(&num_hdrs))
1337 case FLOW_DISSECT_RET_CONTINUE:
1338 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1340 case FLOW_DISSECT_RET_OUT_BAD:
1346 fdret = FLOW_DISSECT_RET_CONTINUE;
1350 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1351 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1355 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1356 target_container, data,
1357 &proto, &nhoff, &hlen, flags);
1361 case NEXTHDR_ROUTING:
1362 case NEXTHDR_DEST: {
1363 u8 _opthdr[2], *opthdr;
1365 if (proto != htons(ETH_P_IPV6))
1368 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1369 data, hlen, &_opthdr);
1371 fdret = FLOW_DISSECT_RET_OUT_BAD;
1375 ip_proto = opthdr[0];
1376 nhoff += (opthdr[1] + 1) << 3;
1378 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1381 case NEXTHDR_FRAGMENT: {
1382 struct frag_hdr _fh, *fh;
1384 if (proto != htons(ETH_P_IPV6))
1387 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1391 fdret = FLOW_DISSECT_RET_OUT_BAD;
1395 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1397 nhoff += sizeof(_fh);
1398 ip_proto = fh->nexthdr;
1400 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1401 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1402 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1403 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1408 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1412 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1413 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1417 proto = htons(ETH_P_IP);
1419 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1420 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1421 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1425 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1429 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1430 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1434 proto = htons(ETH_P_IPV6);
1436 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1437 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1438 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1442 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1447 proto = htons(ETH_P_MPLS_UC);
1448 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1452 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1457 case IPPROTO_ICMPV6:
1458 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1466 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1467 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1468 data, nhoff, ip_proto, hlen);
1470 /* Process result of IP proto processing */
1472 case FLOW_DISSECT_RET_PROTO_AGAIN:
1473 if (skb_flow_dissect_allowed(&num_hdrs))
1476 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1477 if (skb_flow_dissect_allowed(&num_hdrs))
1478 goto ip_proto_again;
1480 case FLOW_DISSECT_RET_OUT_GOOD:
1481 case FLOW_DISSECT_RET_CONTINUE:
1483 case FLOW_DISSECT_RET_OUT_BAD:
1492 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1493 key_basic->n_proto = proto;
1494 key_basic->ip_proto = ip_proto;
1502 EXPORT_SYMBOL(__skb_flow_dissect);
1504 static siphash_aligned_key_t hashrnd;
1505 static __always_inline void __flow_hash_secret_init(void)
1507 net_get_random_once(&hashrnd, sizeof(hashrnd));
1510 static const void *flow_keys_hash_start(const struct flow_keys *flow)
1512 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1513 return &flow->FLOW_KEYS_HASH_START_FIELD;
1516 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1518 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1520 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1522 switch (flow->control.addr_type) {
1523 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1524 diff -= sizeof(flow->addrs.v4addrs);
1526 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1527 diff -= sizeof(flow->addrs.v6addrs);
1529 case FLOW_DISSECTOR_KEY_TIPC:
1530 diff -= sizeof(flow->addrs.tipckey);
1533 return sizeof(*flow) - diff;
1536 __be32 flow_get_u32_src(const struct flow_keys *flow)
1538 switch (flow->control.addr_type) {
1539 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1540 return flow->addrs.v4addrs.src;
1541 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1542 return (__force __be32)ipv6_addr_hash(
1543 &flow->addrs.v6addrs.src);
1544 case FLOW_DISSECTOR_KEY_TIPC:
1545 return flow->addrs.tipckey.key;
1550 EXPORT_SYMBOL(flow_get_u32_src);
1552 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1554 switch (flow->control.addr_type) {
1555 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1556 return flow->addrs.v4addrs.dst;
1557 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1558 return (__force __be32)ipv6_addr_hash(
1559 &flow->addrs.v6addrs.dst);
1564 EXPORT_SYMBOL(flow_get_u32_dst);
1566 /* Sort the source and destination IP and the ports,
1567 * to have consistent hash within the two directions
1569 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1573 switch (keys->control.addr_type) {
1574 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1575 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1576 (__force u32)keys->addrs.v4addrs.src;
1578 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1580 if ((__force u16)keys->ports.dst <
1581 (__force u16)keys->ports.src) {
1582 swap(keys->ports.src, keys->ports.dst);
1585 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1586 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1587 &keys->addrs.v6addrs.src,
1588 sizeof(keys->addrs.v6addrs.dst));
1589 if (addr_diff < 0) {
1590 for (i = 0; i < 4; i++)
1591 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1592 keys->addrs.v6addrs.dst.s6_addr32[i]);
1594 if ((__force u16)keys->ports.dst <
1595 (__force u16)keys->ports.src) {
1596 swap(keys->ports.src, keys->ports.dst);
1602 static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1603 const siphash_key_t *keyval)
1607 __flow_hash_consistentify(keys);
1609 hash = siphash(flow_keys_hash_start(keys),
1610 flow_keys_hash_length(keys), keyval);
1617 u32 flow_hash_from_keys(struct flow_keys *keys)
1619 __flow_hash_secret_init();
1620 return __flow_hash_from_keys(keys, &hashrnd);
1622 EXPORT_SYMBOL(flow_hash_from_keys);
1624 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1625 struct flow_keys *keys,
1626 const siphash_key_t *keyval)
1628 skb_flow_dissect_flow_keys(skb, keys,
1629 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1631 return __flow_hash_from_keys(keys, keyval);
1634 struct _flow_keys_digest_data {
1643 void make_flow_keys_digest(struct flow_keys_digest *digest,
1644 const struct flow_keys *flow)
1646 struct _flow_keys_digest_data *data =
1647 (struct _flow_keys_digest_data *)digest;
1649 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1651 memset(digest, 0, sizeof(*digest));
1653 data->n_proto = flow->basic.n_proto;
1654 data->ip_proto = flow->basic.ip_proto;
1655 data->ports = flow->ports.ports;
1656 data->src = flow->addrs.v4addrs.src;
1657 data->dst = flow->addrs.v4addrs.dst;
1659 EXPORT_SYMBOL(make_flow_keys_digest);
1661 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1663 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1665 struct flow_keys keys;
1667 __flow_hash_secret_init();
1669 memset(&keys, 0, sizeof(keys));
1670 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1671 &keys, NULL, 0, 0, 0,
1672 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1674 return __flow_hash_from_keys(&keys, &hashrnd);
1676 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1679 * __skb_get_hash: calculate a flow hash
1680 * @skb: sk_buff to calculate flow hash from
1682 * This function calculates a flow hash based on src/dst addresses
1683 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1684 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1685 * if hash is a canonical 4-tuple hash over transport ports.
1687 void __skb_get_hash(struct sk_buff *skb)
1689 struct flow_keys keys;
1692 __flow_hash_secret_init();
1694 hash = ___skb_get_hash(skb, &keys, &hashrnd);
1696 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1698 EXPORT_SYMBOL(__skb_get_hash);
1700 __u32 skb_get_hash_perturb(const struct sk_buff *skb,
1701 const siphash_key_t *perturb)
1703 struct flow_keys keys;
1705 return ___skb_get_hash(skb, &keys, perturb);
1707 EXPORT_SYMBOL(skb_get_hash_perturb);
1709 u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1710 const struct flow_keys_basic *keys, int hlen)
1712 u32 poff = keys->control.thoff;
1714 /* skip L4 headers for fragments after the first */
1715 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1716 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1719 switch (keys->basic.ip_proto) {
1721 /* access doff as u8 to avoid unaligned access */
1725 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1726 data, hlen, &_doff);
1730 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1734 case IPPROTO_UDPLITE:
1735 poff += sizeof(struct udphdr);
1737 /* For the rest, we do not really care about header
1738 * extensions at this point for now.
1741 poff += sizeof(struct icmphdr);
1743 case IPPROTO_ICMPV6:
1744 poff += sizeof(struct icmp6hdr);
1747 poff += sizeof(struct igmphdr);
1750 poff += sizeof(struct dccp_hdr);
1753 poff += sizeof(struct sctphdr);
1761 * skb_get_poff - get the offset to the payload
1762 * @skb: sk_buff to get the payload offset from
1764 * The function will get the offset to the payload as far as it could
1765 * be dissected. The main user is currently BPF, so that we can dynamically
1766 * truncate packets without needing to push actual payload to the user
1767 * space and can analyze headers only, instead.
1769 u32 skb_get_poff(const struct sk_buff *skb)
1771 struct flow_keys_basic keys;
1773 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1777 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1780 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1782 memset(keys, 0, sizeof(*keys));
1784 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1785 sizeof(keys->addrs.v6addrs.src));
1786 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1787 sizeof(keys->addrs.v6addrs.dst));
1788 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1789 keys->ports.src = fl6->fl6_sport;
1790 keys->ports.dst = fl6->fl6_dport;
1791 keys->keyid.keyid = fl6->fl6_gre_key;
1792 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1793 keys->basic.ip_proto = fl6->flowi6_proto;
1795 return flow_hash_from_keys(keys);
1797 EXPORT_SYMBOL(__get_hash_from_flowi6);
1799 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1801 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1802 .offset = offsetof(struct flow_keys, control),
1805 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1806 .offset = offsetof(struct flow_keys, basic),
1809 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1810 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1813 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1814 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1817 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1818 .offset = offsetof(struct flow_keys, addrs.tipckey),
1821 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1822 .offset = offsetof(struct flow_keys, ports),
1825 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1826 .offset = offsetof(struct flow_keys, vlan),
1829 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1830 .offset = offsetof(struct flow_keys, tags),
1833 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1834 .offset = offsetof(struct flow_keys, keyid),
1838 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1840 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1841 .offset = offsetof(struct flow_keys, control),
1844 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1845 .offset = offsetof(struct flow_keys, basic),
1848 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1849 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1852 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1853 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1856 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1857 .offset = offsetof(struct flow_keys, ports),
1861 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1863 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1864 .offset = offsetof(struct flow_keys, control),
1867 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1868 .offset = offsetof(struct flow_keys, basic),
1872 struct flow_dissector flow_keys_dissector __read_mostly;
1873 EXPORT_SYMBOL(flow_keys_dissector);
1875 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1876 EXPORT_SYMBOL(flow_keys_basic_dissector);
1878 static int __init init_default_flow_dissectors(void)
1880 skb_flow_dissector_init(&flow_keys_dissector,
1881 flow_keys_dissector_keys,
1882 ARRAY_SIZE(flow_keys_dissector_keys));
1883 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1884 flow_keys_dissector_symmetric_keys,
1885 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1886 skb_flow_dissector_init(&flow_keys_basic_dissector,
1887 flow_keys_basic_dissector_keys,
1888 ARRAY_SIZE(flow_keys_basic_dissector_keys));
1891 core_initcall(init_default_flow_dissectors);