1 #include <linux/kernel.h>
2 #include <linux/skbuff.h>
3 #include <linux/export.h>
5 #include <linux/ipv6.h>
6 #include <linux/if_vlan.h>
9 #include <linux/igmp.h>
10 #include <linux/icmp.h>
11 #include <linux/sctp.h>
12 #include <linux/dccp.h>
13 #include <linux/if_tunnel.h>
14 #include <linux/if_pppox.h>
15 #include <linux/ppp_defs.h>
16 #include <linux/stddef.h>
17 #include <linux/if_ether.h>
18 #include <linux/mpls.h>
19 #include <net/flow_dissector.h>
20 #include <scsi/fc/fc_fcoe.h>
22 static bool skb_flow_dissector_uses_key(struct flow_dissector *flow_dissector,
23 enum flow_dissector_key_id key_id)
25 return flow_dissector->used_keys & (1 << key_id);
28 static void skb_flow_dissector_set_key(struct flow_dissector *flow_dissector,
29 enum flow_dissector_key_id key_id)
31 flow_dissector->used_keys |= (1 << key_id);
34 static void *skb_flow_dissector_target(struct flow_dissector *flow_dissector,
35 enum flow_dissector_key_id key_id,
36 void *target_container)
38 return ((char *) target_container) + flow_dissector->offset[key_id];
41 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
42 const struct flow_dissector_key *key,
43 unsigned int key_count)
47 memset(flow_dissector, 0, sizeof(*flow_dissector));
49 for (i = 0; i < key_count; i++, key++) {
50 /* User should make sure that every key target offset is withing
51 * boundaries of unsigned short.
53 BUG_ON(key->offset > USHRT_MAX);
54 BUG_ON(skb_flow_dissector_uses_key(flow_dissector,
57 skb_flow_dissector_set_key(flow_dissector, key->key_id);
58 flow_dissector->offset[key->key_id] = key->offset;
61 /* Ensure that the dissector always includes control and basic key.
62 * That way we are able to avoid handling lack of these in fast path.
64 BUG_ON(!skb_flow_dissector_uses_key(flow_dissector,
65 FLOW_DISSECTOR_KEY_CONTROL));
66 BUG_ON(!skb_flow_dissector_uses_key(flow_dissector,
67 FLOW_DISSECTOR_KEY_BASIC));
69 EXPORT_SYMBOL(skb_flow_dissector_init);
72 * __skb_flow_get_ports - extract the upper layer ports and return them
73 * @skb: sk_buff to extract the ports from
74 * @thoff: transport header offset
75 * @ip_proto: protocol for which to get port offset
76 * @data: raw buffer pointer to the packet, if NULL use skb->data
77 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
79 * The function will try to retrieve the ports at offset thoff + poff where poff
80 * is the protocol port offset returned from proto_ports_offset
82 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
85 int poff = proto_ports_offset(ip_proto);
89 hlen = skb_headlen(skb);
93 __be32 *ports, _ports;
95 ports = __skb_header_pointer(skb, thoff + poff,
96 sizeof(_ports), data, hlen, &_ports);
103 EXPORT_SYMBOL(__skb_flow_get_ports);
106 * __skb_flow_dissect - extract the flow_keys struct and return it
107 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
108 * @flow_dissector: list of keys to dissect
109 * @target_container: target structure to put dissected values into
110 * @data: raw buffer pointer to the packet, if NULL use skb->data
111 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
112 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
113 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
115 * The function will try to retrieve individual keys into target specified
116 * by flow_dissector from either the skbuff or a raw buffer specified by the
119 * Caller must take care of zeroing target container memory.
121 bool __skb_flow_dissect(const struct sk_buff *skb,
122 struct flow_dissector *flow_dissector,
123 void *target_container,
124 void *data, __be16 proto, int nhoff, int hlen)
126 struct flow_dissector_key_control *key_control;
127 struct flow_dissector_key_basic *key_basic;
128 struct flow_dissector_key_addrs *key_addrs;
129 struct flow_dissector_key_ports *key_ports;
130 struct flow_dissector_key_tags *key_tags;
131 struct flow_dissector_key_keyid *key_keyid;
137 proto = skb->protocol;
138 nhoff = skb_network_offset(skb);
139 hlen = skb_headlen(skb);
142 /* It is ensured by skb_flow_dissector_init() that control key will
145 key_control = skb_flow_dissector_target(flow_dissector,
146 FLOW_DISSECTOR_KEY_CONTROL,
149 /* It is ensured by skb_flow_dissector_init() that basic key will
152 key_basic = skb_flow_dissector_target(flow_dissector,
153 FLOW_DISSECTOR_KEY_BASIC,
156 if (skb_flow_dissector_uses_key(flow_dissector,
157 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
158 struct ethhdr *eth = eth_hdr(skb);
159 struct flow_dissector_key_eth_addrs *key_eth_addrs;
161 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
162 FLOW_DISSECTOR_KEY_ETH_ADDRS,
164 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs));
169 case htons(ETH_P_IP): {
170 const struct iphdr *iph;
173 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
174 if (!iph || iph->ihl < 5)
176 nhoff += iph->ihl * 4;
178 ip_proto = iph->protocol;
179 if (ip_is_fragment(iph))
182 if (!skb_flow_dissector_uses_key(flow_dissector,
183 FLOW_DISSECTOR_KEY_IPV4_ADDRS))
186 key_addrs = skb_flow_dissector_target(flow_dissector,
187 FLOW_DISSECTOR_KEY_IPV4_ADDRS, target_container);
188 memcpy(&key_addrs->v4addrs, &iph->saddr,
189 sizeof(key_addrs->v4addrs));
190 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
193 case htons(ETH_P_IPV6): {
194 const struct ipv6hdr *iph;
199 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
203 ip_proto = iph->nexthdr;
204 nhoff += sizeof(struct ipv6hdr);
206 if (skb_flow_dissector_uses_key(flow_dissector,
207 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
208 struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs;
210 key_ipv6_addrs = skb_flow_dissector_target(flow_dissector,
211 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
214 memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs));
215 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
218 flow_label = ip6_flowlabel(iph);
220 if (skb_flow_dissector_uses_key(flow_dissector,
221 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
222 key_tags = skb_flow_dissector_target(flow_dissector,
223 FLOW_DISSECTOR_KEY_FLOW_LABEL,
225 key_tags->flow_label = ntohl(flow_label);
231 case htons(ETH_P_8021AD):
232 case htons(ETH_P_8021Q): {
233 const struct vlan_hdr *vlan;
234 struct vlan_hdr _vlan;
236 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan);
240 if (skb_flow_dissector_uses_key(flow_dissector,
241 FLOW_DISSECTOR_KEY_VLANID)) {
242 key_tags = skb_flow_dissector_target(flow_dissector,
243 FLOW_DISSECTOR_KEY_VLANID,
246 key_tags->vlan_id = skb_vlan_tag_get_id(skb);
249 proto = vlan->h_vlan_encapsulated_proto;
250 nhoff += sizeof(*vlan);
253 case htons(ETH_P_PPP_SES): {
255 struct pppoe_hdr hdr;
258 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
262 nhoff += PPPOE_SES_HLEN;
266 case htons(PPP_IPV6):
272 case htons(ETH_P_TIPC): {
277 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
281 if (skb_flow_dissector_uses_key(flow_dissector,
282 FLOW_DISSECTOR_KEY_TIPC_ADDRS)) {
283 key_addrs = skb_flow_dissector_target(flow_dissector,
284 FLOW_DISSECTOR_KEY_TIPC_ADDRS,
286 key_addrs->tipcaddrs.srcnode = hdr->srcnode;
287 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS;
292 case htons(ETH_P_MPLS_UC):
293 case htons(ETH_P_MPLS_MC): {
294 struct mpls_label *hdr, _hdr[2];
296 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
301 if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >>
302 MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) {
303 if (skb_flow_dissector_uses_key(flow_dissector,
304 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
305 key_keyid = skb_flow_dissector_target(flow_dissector,
306 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
308 key_keyid->keyid = hdr[1].entry &
309 htonl(MPLS_LS_LABEL_MASK);
318 case htons(ETH_P_FCOE):
319 key_control->thoff = (u16)(nhoff + FCOE_HEADER_LEN);
333 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
337 * Only look inside GRE if version zero and no
340 if (hdr->flags & (GRE_VERSION | GRE_ROUTING))
345 if (hdr->flags & GRE_CSUM)
347 if (hdr->flags & GRE_KEY) {
351 keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid),
352 data, hlen, &_keyid);
357 if (skb_flow_dissector_uses_key(flow_dissector,
358 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
359 key_keyid = skb_flow_dissector_target(flow_dissector,
360 FLOW_DISSECTOR_KEY_GRE_KEYID,
362 key_keyid->keyid = *keyid;
366 if (hdr->flags & GRE_SEQ)
368 if (proto == htons(ETH_P_TEB)) {
369 const struct ethhdr *eth;
372 eth = __skb_header_pointer(skb, nhoff,
377 proto = eth->h_proto;
378 nhoff += sizeof(*eth);
383 case NEXTHDR_ROUTING:
385 u8 _opthdr[2], *opthdr;
387 if (proto != htons(ETH_P_IPV6))
390 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
391 data, hlen, &_opthdr);
395 ip_proto = opthdr[0];
396 nhoff += (opthdr[1] + 1) << 3;
401 proto = htons(ETH_P_IP);
404 proto = htons(ETH_P_IPV6);
407 proto = htons(ETH_P_MPLS_UC);
413 if (skb_flow_dissector_uses_key(flow_dissector,
414 FLOW_DISSECTOR_KEY_PORTS)) {
415 key_ports = skb_flow_dissector_target(flow_dissector,
416 FLOW_DISSECTOR_KEY_PORTS,
418 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
426 key_basic->n_proto = proto;
427 key_basic->ip_proto = ip_proto;
428 key_control->thoff = (u16)nhoff;
432 EXPORT_SYMBOL(__skb_flow_dissect);
434 static u32 hashrnd __read_mostly;
435 static __always_inline void __flow_hash_secret_init(void)
437 net_get_random_once(&hashrnd, sizeof(hashrnd));
440 static __always_inline u32 __flow_hash_words(u32 *words, u32 length, u32 keyval)
442 return jhash2(words, length, keyval);
445 static inline void *flow_keys_hash_start(struct flow_keys *flow)
447 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
448 return (void *)flow + FLOW_KEYS_HASH_OFFSET;
451 static inline size_t flow_keys_hash_length(struct flow_keys *flow)
453 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
454 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
455 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
456 sizeof(*flow) - sizeof(flow->addrs));
458 switch (flow->control.addr_type) {
459 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
460 diff -= sizeof(flow->addrs.v4addrs);
462 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
463 diff -= sizeof(flow->addrs.v6addrs);
465 case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
466 diff -= sizeof(flow->addrs.tipcaddrs);
469 return (sizeof(*flow) - diff) / sizeof(u32);
472 __be32 flow_get_u32_src(const struct flow_keys *flow)
474 switch (flow->control.addr_type) {
475 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
476 return flow->addrs.v4addrs.src;
477 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
478 return (__force __be32)ipv6_addr_hash(
479 &flow->addrs.v6addrs.src);
480 case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
481 return flow->addrs.tipcaddrs.srcnode;
486 EXPORT_SYMBOL(flow_get_u32_src);
488 __be32 flow_get_u32_dst(const struct flow_keys *flow)
490 switch (flow->control.addr_type) {
491 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
492 return flow->addrs.v4addrs.dst;
493 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
494 return (__force __be32)ipv6_addr_hash(
495 &flow->addrs.v6addrs.dst);
500 EXPORT_SYMBOL(flow_get_u32_dst);
502 static inline void __flow_hash_consistentify(struct flow_keys *keys)
506 switch (keys->control.addr_type) {
507 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
508 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
509 (__force u32)keys->addrs.v4addrs.src;
510 if ((addr_diff < 0) ||
512 ((__force u16)keys->ports.dst <
513 (__force u16)keys->ports.src))) {
514 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
515 swap(keys->ports.src, keys->ports.dst);
518 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
519 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
520 &keys->addrs.v6addrs.src,
521 sizeof(keys->addrs.v6addrs.dst));
522 if ((addr_diff < 0) ||
524 ((__force u16)keys->ports.dst <
525 (__force u16)keys->ports.src))) {
526 for (i = 0; i < 4; i++)
527 swap(keys->addrs.v6addrs.src.s6_addr32[i],
528 keys->addrs.v6addrs.dst.s6_addr32[i]);
529 swap(keys->ports.src, keys->ports.dst);
535 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
539 __flow_hash_consistentify(keys);
541 hash = __flow_hash_words((u32 *)flow_keys_hash_start(keys),
542 flow_keys_hash_length(keys), keyval);
549 u32 flow_hash_from_keys(struct flow_keys *keys)
551 __flow_hash_secret_init();
552 return __flow_hash_from_keys(keys, hashrnd);
554 EXPORT_SYMBOL(flow_hash_from_keys);
556 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
557 struct flow_keys *keys, u32 keyval)
559 if (!skb_flow_dissect_flow_keys(skb, keys))
562 return __flow_hash_from_keys(keys, keyval);
565 struct _flow_keys_digest_data {
574 void make_flow_keys_digest(struct flow_keys_digest *digest,
575 const struct flow_keys *flow)
577 struct _flow_keys_digest_data *data =
578 (struct _flow_keys_digest_data *)digest;
580 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
582 memset(digest, 0, sizeof(*digest));
584 data->n_proto = flow->basic.n_proto;
585 data->ip_proto = flow->basic.ip_proto;
586 data->ports = flow->ports.ports;
587 data->src = flow->addrs.v4addrs.src;
588 data->dst = flow->addrs.v4addrs.dst;
590 EXPORT_SYMBOL(make_flow_keys_digest);
593 * __skb_get_hash: calculate a flow hash
594 * @skb: sk_buff to calculate flow hash from
596 * This function calculates a flow hash based on src/dst addresses
597 * and src/dst port numbers. Sets hash in skb to non-zero hash value
598 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
599 * if hash is a canonical 4-tuple hash over transport ports.
601 void __skb_get_hash(struct sk_buff *skb)
603 struct flow_keys keys;
606 __flow_hash_secret_init();
608 hash = ___skb_get_hash(skb, &keys, hashrnd);
612 __skb_set_sw_hash(skb, hash,
613 flow_keys_have_l4(&keys));
615 EXPORT_SYMBOL(__skb_get_hash);
617 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
619 struct flow_keys keys;
621 return ___skb_get_hash(skb, &keys, perturb);
623 EXPORT_SYMBOL(skb_get_hash_perturb);
625 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, struct flowi6 *fl6)
627 struct flow_keys keys;
629 memset(&keys, 0, sizeof(keys));
631 memcpy(&keys.addrs.v6addrs.src, &fl6->saddr,
632 sizeof(keys.addrs.v6addrs.src));
633 memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr,
634 sizeof(keys.addrs.v6addrs.dst));
635 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
636 keys.ports.src = fl6->fl6_sport;
637 keys.ports.dst = fl6->fl6_dport;
638 keys.keyid.keyid = fl6->fl6_gre_key;
639 keys.tags.flow_label = (__force u32)fl6->flowlabel;
640 keys.basic.ip_proto = fl6->flowi6_proto;
642 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
643 flow_keys_have_l4(&keys));
647 EXPORT_SYMBOL(__skb_get_hash_flowi6);
649 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, struct flowi4 *fl4)
651 struct flow_keys keys;
653 memset(&keys, 0, sizeof(keys));
655 keys.addrs.v4addrs.src = fl4->saddr;
656 keys.addrs.v4addrs.dst = fl4->daddr;
657 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
658 keys.ports.src = fl4->fl4_sport;
659 keys.ports.dst = fl4->fl4_dport;
660 keys.keyid.keyid = fl4->fl4_gre_key;
661 keys.basic.ip_proto = fl4->flowi4_proto;
663 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
664 flow_keys_have_l4(&keys));
668 EXPORT_SYMBOL(__skb_get_hash_flowi4);
670 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
671 const struct flow_keys *keys, int hlen)
673 u32 poff = keys->control.thoff;
675 switch (keys->basic.ip_proto) {
677 /* access doff as u8 to avoid unaligned access */
681 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
686 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
690 case IPPROTO_UDPLITE:
691 poff += sizeof(struct udphdr);
693 /* For the rest, we do not really care about header
694 * extensions at this point for now.
697 poff += sizeof(struct icmphdr);
700 poff += sizeof(struct icmp6hdr);
703 poff += sizeof(struct igmphdr);
706 poff += sizeof(struct dccp_hdr);
709 poff += sizeof(struct sctphdr);
717 * skb_get_poff - get the offset to the payload
718 * @skb: sk_buff to get the payload offset from
720 * The function will get the offset to the payload as far as it could
721 * be dissected. The main user is currently BPF, so that we can dynamically
722 * truncate packets without needing to push actual payload to the user
723 * space and can analyze headers only, instead.
725 u32 skb_get_poff(const struct sk_buff *skb)
727 struct flow_keys keys;
729 if (!skb_flow_dissect_flow_keys(skb, &keys))
732 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
735 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
737 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
738 .offset = offsetof(struct flow_keys, control),
741 .key_id = FLOW_DISSECTOR_KEY_BASIC,
742 .offset = offsetof(struct flow_keys, basic),
745 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
746 .offset = offsetof(struct flow_keys, addrs.v4addrs),
749 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
750 .offset = offsetof(struct flow_keys, addrs.v6addrs),
753 .key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS,
754 .offset = offsetof(struct flow_keys, addrs.tipcaddrs),
757 .key_id = FLOW_DISSECTOR_KEY_PORTS,
758 .offset = offsetof(struct flow_keys, ports),
761 .key_id = FLOW_DISSECTOR_KEY_VLANID,
762 .offset = offsetof(struct flow_keys, tags),
765 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
766 .offset = offsetof(struct flow_keys, tags),
769 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
770 .offset = offsetof(struct flow_keys, keyid),
774 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
776 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
777 .offset = offsetof(struct flow_keys, control),
780 .key_id = FLOW_DISSECTOR_KEY_BASIC,
781 .offset = offsetof(struct flow_keys, basic),
785 struct flow_dissector flow_keys_dissector __read_mostly;
786 EXPORT_SYMBOL(flow_keys_dissector);
788 struct flow_dissector flow_keys_buf_dissector __read_mostly;
790 static int __init init_default_flow_dissectors(void)
792 skb_flow_dissector_init(&flow_keys_dissector,
793 flow_keys_dissector_keys,
794 ARRAY_SIZE(flow_keys_dissector_keys));
795 skb_flow_dissector_init(&flow_keys_buf_dissector,
796 flow_keys_buf_dissector_keys,
797 ARRAY_SIZE(flow_keys_buf_dissector_keys));
801 late_initcall_sync(init_default_flow_dissectors);