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,
127 struct flow_dissector_key_control *key_control;
128 struct flow_dissector_key_basic *key_basic;
129 struct flow_dissector_key_addrs *key_addrs;
130 struct flow_dissector_key_ports *key_ports;
131 struct flow_dissector_key_tags *key_tags;
132 struct flow_dissector_key_keyid *key_keyid;
138 proto = skb->protocol;
139 nhoff = skb_network_offset(skb);
140 hlen = skb_headlen(skb);
143 /* It is ensured by skb_flow_dissector_init() that control key will
146 key_control = skb_flow_dissector_target(flow_dissector,
147 FLOW_DISSECTOR_KEY_CONTROL,
150 /* It is ensured by skb_flow_dissector_init() that basic key will
153 key_basic = skb_flow_dissector_target(flow_dissector,
154 FLOW_DISSECTOR_KEY_BASIC,
157 if (skb_flow_dissector_uses_key(flow_dissector,
158 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
159 struct ethhdr *eth = eth_hdr(skb);
160 struct flow_dissector_key_eth_addrs *key_eth_addrs;
162 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
163 FLOW_DISSECTOR_KEY_ETH_ADDRS,
165 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs));
170 case htons(ETH_P_IP): {
171 const struct iphdr *iph;
174 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
175 if (!iph || iph->ihl < 5)
177 nhoff += iph->ihl * 4;
179 ip_proto = iph->protocol;
181 if (!skb_flow_dissector_uses_key(flow_dissector,
182 FLOW_DISSECTOR_KEY_IPV4_ADDRS))
185 key_addrs = skb_flow_dissector_target(flow_dissector,
186 FLOW_DISSECTOR_KEY_IPV4_ADDRS, target_container);
187 memcpy(&key_addrs->v4addrs, &iph->saddr,
188 sizeof(key_addrs->v4addrs));
189 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
191 if (ip_is_fragment(iph)) {
192 key_control->is_fragment = 1;
194 if (iph->frag_off & htons(IP_OFFSET)) {
197 key_control->first_frag = 1;
198 if (!(flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG))
205 case htons(ETH_P_IPV6): {
206 const struct ipv6hdr *iph;
211 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
215 ip_proto = iph->nexthdr;
216 nhoff += sizeof(struct ipv6hdr);
218 if (skb_flow_dissector_uses_key(flow_dissector,
219 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
220 struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs;
222 key_ipv6_addrs = skb_flow_dissector_target(flow_dissector,
223 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
226 memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs));
227 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
230 flow_label = ip6_flowlabel(iph);
232 if (skb_flow_dissector_uses_key(flow_dissector,
233 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
234 key_tags = skb_flow_dissector_target(flow_dissector,
235 FLOW_DISSECTOR_KEY_FLOW_LABEL,
237 key_tags->flow_label = ntohl(flow_label);
243 case htons(ETH_P_8021AD):
244 case htons(ETH_P_8021Q): {
245 const struct vlan_hdr *vlan;
246 struct vlan_hdr _vlan;
248 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan);
252 if (skb_flow_dissector_uses_key(flow_dissector,
253 FLOW_DISSECTOR_KEY_VLANID)) {
254 key_tags = skb_flow_dissector_target(flow_dissector,
255 FLOW_DISSECTOR_KEY_VLANID,
258 key_tags->vlan_id = skb_vlan_tag_get_id(skb);
261 proto = vlan->h_vlan_encapsulated_proto;
262 nhoff += sizeof(*vlan);
265 case htons(ETH_P_PPP_SES): {
267 struct pppoe_hdr hdr;
270 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
274 nhoff += PPPOE_SES_HLEN;
278 case htons(PPP_IPV6):
284 case htons(ETH_P_TIPC): {
289 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
293 if (skb_flow_dissector_uses_key(flow_dissector,
294 FLOW_DISSECTOR_KEY_TIPC_ADDRS)) {
295 key_addrs = skb_flow_dissector_target(flow_dissector,
296 FLOW_DISSECTOR_KEY_TIPC_ADDRS,
298 key_addrs->tipcaddrs.srcnode = hdr->srcnode;
299 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS;
304 case htons(ETH_P_MPLS_UC):
305 case htons(ETH_P_MPLS_MC): {
306 struct mpls_label *hdr, _hdr[2];
308 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
313 if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >>
314 MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) {
315 if (skb_flow_dissector_uses_key(flow_dissector,
316 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
317 key_keyid = skb_flow_dissector_target(flow_dissector,
318 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
320 key_keyid->keyid = hdr[1].entry &
321 htonl(MPLS_LS_LABEL_MASK);
330 case htons(ETH_P_FCOE):
331 key_control->thoff = (u16)(nhoff + FCOE_HEADER_LEN);
345 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
349 * Only look inside GRE if version zero and no
352 if (hdr->flags & (GRE_VERSION | GRE_ROUTING))
357 if (hdr->flags & GRE_CSUM)
359 if (hdr->flags & GRE_KEY) {
363 keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid),
364 data, hlen, &_keyid);
369 if (skb_flow_dissector_uses_key(flow_dissector,
370 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
371 key_keyid = skb_flow_dissector_target(flow_dissector,
372 FLOW_DISSECTOR_KEY_GRE_KEYID,
374 key_keyid->keyid = *keyid;
378 if (hdr->flags & GRE_SEQ)
380 if (proto == htons(ETH_P_TEB)) {
381 const struct ethhdr *eth;
384 eth = __skb_header_pointer(skb, nhoff,
389 proto = eth->h_proto;
390 nhoff += sizeof(*eth);
395 case NEXTHDR_ROUTING:
397 u8 _opthdr[2], *opthdr;
399 if (proto != htons(ETH_P_IPV6))
402 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
403 data, hlen, &_opthdr);
407 ip_proto = opthdr[0];
408 nhoff += (opthdr[1] + 1) << 3;
413 proto = htons(ETH_P_IP);
416 proto = htons(ETH_P_IPV6);
419 proto = htons(ETH_P_MPLS_UC);
425 if (skb_flow_dissector_uses_key(flow_dissector,
426 FLOW_DISSECTOR_KEY_PORTS)) {
427 key_ports = skb_flow_dissector_target(flow_dissector,
428 FLOW_DISSECTOR_KEY_PORTS,
430 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
438 key_basic->n_proto = proto;
439 key_basic->ip_proto = ip_proto;
440 key_control->thoff = (u16)nhoff;
444 EXPORT_SYMBOL(__skb_flow_dissect);
446 static u32 hashrnd __read_mostly;
447 static __always_inline void __flow_hash_secret_init(void)
449 net_get_random_once(&hashrnd, sizeof(hashrnd));
452 static __always_inline u32 __flow_hash_words(u32 *words, u32 length, u32 keyval)
454 return jhash2(words, length, keyval);
457 static inline void *flow_keys_hash_start(struct flow_keys *flow)
459 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
460 return (void *)flow + FLOW_KEYS_HASH_OFFSET;
463 static inline size_t flow_keys_hash_length(struct flow_keys *flow)
465 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
466 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
467 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
468 sizeof(*flow) - sizeof(flow->addrs));
470 switch (flow->control.addr_type) {
471 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
472 diff -= sizeof(flow->addrs.v4addrs);
474 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
475 diff -= sizeof(flow->addrs.v6addrs);
477 case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
478 diff -= sizeof(flow->addrs.tipcaddrs);
481 return (sizeof(*flow) - diff) / sizeof(u32);
484 __be32 flow_get_u32_src(const struct flow_keys *flow)
486 switch (flow->control.addr_type) {
487 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
488 return flow->addrs.v4addrs.src;
489 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
490 return (__force __be32)ipv6_addr_hash(
491 &flow->addrs.v6addrs.src);
492 case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
493 return flow->addrs.tipcaddrs.srcnode;
498 EXPORT_SYMBOL(flow_get_u32_src);
500 __be32 flow_get_u32_dst(const struct flow_keys *flow)
502 switch (flow->control.addr_type) {
503 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
504 return flow->addrs.v4addrs.dst;
505 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
506 return (__force __be32)ipv6_addr_hash(
507 &flow->addrs.v6addrs.dst);
512 EXPORT_SYMBOL(flow_get_u32_dst);
514 static inline void __flow_hash_consistentify(struct flow_keys *keys)
518 switch (keys->control.addr_type) {
519 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
520 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
521 (__force u32)keys->addrs.v4addrs.src;
522 if ((addr_diff < 0) ||
524 ((__force u16)keys->ports.dst <
525 (__force u16)keys->ports.src))) {
526 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
527 swap(keys->ports.src, keys->ports.dst);
530 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
531 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
532 &keys->addrs.v6addrs.src,
533 sizeof(keys->addrs.v6addrs.dst));
534 if ((addr_diff < 0) ||
536 ((__force u16)keys->ports.dst <
537 (__force u16)keys->ports.src))) {
538 for (i = 0; i < 4; i++)
539 swap(keys->addrs.v6addrs.src.s6_addr32[i],
540 keys->addrs.v6addrs.dst.s6_addr32[i]);
541 swap(keys->ports.src, keys->ports.dst);
547 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
551 __flow_hash_consistentify(keys);
553 hash = __flow_hash_words((u32 *)flow_keys_hash_start(keys),
554 flow_keys_hash_length(keys), keyval);
561 u32 flow_hash_from_keys(struct flow_keys *keys)
563 __flow_hash_secret_init();
564 return __flow_hash_from_keys(keys, hashrnd);
566 EXPORT_SYMBOL(flow_hash_from_keys);
568 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
569 struct flow_keys *keys, u32 keyval)
571 if (!skb_flow_dissect_flow_keys(skb, keys, 0))
574 return __flow_hash_from_keys(keys, keyval);
577 struct _flow_keys_digest_data {
586 void make_flow_keys_digest(struct flow_keys_digest *digest,
587 const struct flow_keys *flow)
589 struct _flow_keys_digest_data *data =
590 (struct _flow_keys_digest_data *)digest;
592 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
594 memset(digest, 0, sizeof(*digest));
596 data->n_proto = flow->basic.n_proto;
597 data->ip_proto = flow->basic.ip_proto;
598 data->ports = flow->ports.ports;
599 data->src = flow->addrs.v4addrs.src;
600 data->dst = flow->addrs.v4addrs.dst;
602 EXPORT_SYMBOL(make_flow_keys_digest);
605 * __skb_get_hash: calculate a flow hash
606 * @skb: sk_buff to calculate flow hash from
608 * This function calculates a flow hash based on src/dst addresses
609 * and src/dst port numbers. Sets hash in skb to non-zero hash value
610 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
611 * if hash is a canonical 4-tuple hash over transport ports.
613 void __skb_get_hash(struct sk_buff *skb)
615 struct flow_keys keys;
618 __flow_hash_secret_init();
620 hash = ___skb_get_hash(skb, &keys, hashrnd);
624 __skb_set_sw_hash(skb, hash,
625 flow_keys_have_l4(&keys));
627 EXPORT_SYMBOL(__skb_get_hash);
629 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
631 struct flow_keys keys;
633 return ___skb_get_hash(skb, &keys, perturb);
635 EXPORT_SYMBOL(skb_get_hash_perturb);
637 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, struct flowi6 *fl6)
639 struct flow_keys keys;
641 memset(&keys, 0, sizeof(keys));
643 memcpy(&keys.addrs.v6addrs.src, &fl6->saddr,
644 sizeof(keys.addrs.v6addrs.src));
645 memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr,
646 sizeof(keys.addrs.v6addrs.dst));
647 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
648 keys.ports.src = fl6->fl6_sport;
649 keys.ports.dst = fl6->fl6_dport;
650 keys.keyid.keyid = fl6->fl6_gre_key;
651 keys.tags.flow_label = (__force u32)fl6->flowlabel;
652 keys.basic.ip_proto = fl6->flowi6_proto;
654 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
655 flow_keys_have_l4(&keys));
659 EXPORT_SYMBOL(__skb_get_hash_flowi6);
661 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, struct flowi4 *fl4)
663 struct flow_keys keys;
665 memset(&keys, 0, sizeof(keys));
667 keys.addrs.v4addrs.src = fl4->saddr;
668 keys.addrs.v4addrs.dst = fl4->daddr;
669 keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
670 keys.ports.src = fl4->fl4_sport;
671 keys.ports.dst = fl4->fl4_dport;
672 keys.keyid.keyid = fl4->fl4_gre_key;
673 keys.basic.ip_proto = fl4->flowi4_proto;
675 __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
676 flow_keys_have_l4(&keys));
680 EXPORT_SYMBOL(__skb_get_hash_flowi4);
682 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
683 const struct flow_keys *keys, int hlen)
685 u32 poff = keys->control.thoff;
687 switch (keys->basic.ip_proto) {
689 /* access doff as u8 to avoid unaligned access */
693 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
698 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
702 case IPPROTO_UDPLITE:
703 poff += sizeof(struct udphdr);
705 /* For the rest, we do not really care about header
706 * extensions at this point for now.
709 poff += sizeof(struct icmphdr);
712 poff += sizeof(struct icmp6hdr);
715 poff += sizeof(struct igmphdr);
718 poff += sizeof(struct dccp_hdr);
721 poff += sizeof(struct sctphdr);
729 * skb_get_poff - get the offset to the payload
730 * @skb: sk_buff to get the payload offset from
732 * The function will get the offset to the payload as far as it could
733 * be dissected. The main user is currently BPF, so that we can dynamically
734 * truncate packets without needing to push actual payload to the user
735 * space and can analyze headers only, instead.
737 u32 skb_get_poff(const struct sk_buff *skb)
739 struct flow_keys keys;
741 if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
744 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
747 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
749 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
750 .offset = offsetof(struct flow_keys, control),
753 .key_id = FLOW_DISSECTOR_KEY_BASIC,
754 .offset = offsetof(struct flow_keys, basic),
757 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
758 .offset = offsetof(struct flow_keys, addrs.v4addrs),
761 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
762 .offset = offsetof(struct flow_keys, addrs.v6addrs),
765 .key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS,
766 .offset = offsetof(struct flow_keys, addrs.tipcaddrs),
769 .key_id = FLOW_DISSECTOR_KEY_PORTS,
770 .offset = offsetof(struct flow_keys, ports),
773 .key_id = FLOW_DISSECTOR_KEY_VLANID,
774 .offset = offsetof(struct flow_keys, tags),
777 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
778 .offset = offsetof(struct flow_keys, tags),
781 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
782 .offset = offsetof(struct flow_keys, keyid),
786 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
788 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
789 .offset = offsetof(struct flow_keys, control),
792 .key_id = FLOW_DISSECTOR_KEY_BASIC,
793 .offset = offsetof(struct flow_keys, basic),
797 struct flow_dissector flow_keys_dissector __read_mostly;
798 EXPORT_SYMBOL(flow_keys_dissector);
800 struct flow_dissector flow_keys_buf_dissector __read_mostly;
802 static int __init init_default_flow_dissectors(void)
804 skb_flow_dissector_init(&flow_keys_dissector,
805 flow_keys_dissector_keys,
806 ARRAY_SIZE(flow_keys_dissector_keys));
807 skb_flow_dissector_init(&flow_keys_buf_dissector,
808 flow_keys_buf_dissector_keys,
809 ARRAY_SIZE(flow_keys_buf_dissector_keys));
813 late_initcall_sync(init_default_flow_dissectors);