Loading...
1#include <linux/kernel.h>
2#include <linux/skbuff.h>
3#include <linux/export.h>
4#include <linux/ip.h>
5#include <linux/ipv6.h>
6#include <linux/if_vlan.h>
7#include <net/dsa.h>
8#include <net/dst_metadata.h>
9#include <net/ip.h>
10#include <net/ipv6.h>
11#include <net/gre.h>
12#include <net/pptp.h>
13#include <net/tipc.h>
14#include <linux/igmp.h>
15#include <linux/icmp.h>
16#include <linux/sctp.h>
17#include <linux/dccp.h>
18#include <linux/if_tunnel.h>
19#include <linux/if_pppox.h>
20#include <linux/ppp_defs.h>
21#include <linux/stddef.h>
22#include <linux/if_ether.h>
23#include <linux/mpls.h>
24#include <linux/tcp.h>
25#include <net/flow_dissector.h>
26#include <scsi/fc/fc_fcoe.h>
27#include <uapi/linux/batadv_packet.h>
28
29static void dissector_set_key(struct flow_dissector *flow_dissector,
30 enum flow_dissector_key_id key_id)
31{
32 flow_dissector->used_keys |= (1 << key_id);
33}
34
35void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
36 const struct flow_dissector_key *key,
37 unsigned int key_count)
38{
39 unsigned int i;
40
41 memset(flow_dissector, 0, sizeof(*flow_dissector));
42
43 for (i = 0; i < key_count; i++, key++) {
44 /* User should make sure that every key target offset is withing
45 * boundaries of unsigned short.
46 */
47 BUG_ON(key->offset > USHRT_MAX);
48 BUG_ON(dissector_uses_key(flow_dissector,
49 key->key_id));
50
51 dissector_set_key(flow_dissector, key->key_id);
52 flow_dissector->offset[key->key_id] = key->offset;
53 }
54
55 /* Ensure that the dissector always includes control and basic key.
56 * That way we are able to avoid handling lack of these in fast path.
57 */
58 BUG_ON(!dissector_uses_key(flow_dissector,
59 FLOW_DISSECTOR_KEY_CONTROL));
60 BUG_ON(!dissector_uses_key(flow_dissector,
61 FLOW_DISSECTOR_KEY_BASIC));
62}
63EXPORT_SYMBOL(skb_flow_dissector_init);
64
65/**
66 * skb_flow_get_be16 - extract be16 entity
67 * @skb: sk_buff to extract from
68 * @poff: offset to extract at
69 * @data: raw buffer pointer to the packet
70 * @hlen: packet header length
71 *
72 * The function will try to retrieve a be32 entity at
73 * offset poff
74 */
75static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
76 void *data, int hlen)
77{
78 __be16 *u, _u;
79
80 u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
81 if (u)
82 return *u;
83
84 return 0;
85}
86
87/**
88 * __skb_flow_get_ports - extract the upper layer ports and return them
89 * @skb: sk_buff to extract the ports from
90 * @thoff: transport header offset
91 * @ip_proto: protocol for which to get port offset
92 * @data: raw buffer pointer to the packet, if NULL use skb->data
93 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
94 *
95 * The function will try to retrieve the ports at offset thoff + poff where poff
96 * is the protocol port offset returned from proto_ports_offset
97 */
98__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
99 void *data, int hlen)
100{
101 int poff = proto_ports_offset(ip_proto);
102
103 if (!data) {
104 data = skb->data;
105 hlen = skb_headlen(skb);
106 }
107
108 if (poff >= 0) {
109 __be32 *ports, _ports;
110
111 ports = __skb_header_pointer(skb, thoff + poff,
112 sizeof(_ports), data, hlen, &_ports);
113 if (ports)
114 return *ports;
115 }
116
117 return 0;
118}
119EXPORT_SYMBOL(__skb_flow_get_ports);
120
121static void
122skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
123 struct flow_dissector *flow_dissector,
124 void *target_container)
125{
126 struct flow_dissector_key_control *ctrl;
127
128 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
129 return;
130
131 ctrl = skb_flow_dissector_target(flow_dissector,
132 FLOW_DISSECTOR_KEY_ENC_CONTROL,
133 target_container);
134 ctrl->addr_type = type;
135}
136
137void
138skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
139 struct flow_dissector *flow_dissector,
140 void *target_container)
141{
142 struct ip_tunnel_info *info;
143 struct ip_tunnel_key *key;
144
145 /* A quick check to see if there might be something to do. */
146 if (!dissector_uses_key(flow_dissector,
147 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
148 !dissector_uses_key(flow_dissector,
149 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
150 !dissector_uses_key(flow_dissector,
151 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
152 !dissector_uses_key(flow_dissector,
153 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
154 !dissector_uses_key(flow_dissector,
155 FLOW_DISSECTOR_KEY_ENC_PORTS))
156 return;
157
158 info = skb_tunnel_info(skb);
159 if (!info)
160 return;
161
162 key = &info->key;
163
164 switch (ip_tunnel_info_af(info)) {
165 case AF_INET:
166 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
167 flow_dissector,
168 target_container);
169 if (dissector_uses_key(flow_dissector,
170 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
171 struct flow_dissector_key_ipv4_addrs *ipv4;
172
173 ipv4 = skb_flow_dissector_target(flow_dissector,
174 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
175 target_container);
176 ipv4->src = key->u.ipv4.src;
177 ipv4->dst = key->u.ipv4.dst;
178 }
179 break;
180 case AF_INET6:
181 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
182 flow_dissector,
183 target_container);
184 if (dissector_uses_key(flow_dissector,
185 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
186 struct flow_dissector_key_ipv6_addrs *ipv6;
187
188 ipv6 = skb_flow_dissector_target(flow_dissector,
189 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
190 target_container);
191 ipv6->src = key->u.ipv6.src;
192 ipv6->dst = key->u.ipv6.dst;
193 }
194 break;
195 }
196
197 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
198 struct flow_dissector_key_keyid *keyid;
199
200 keyid = skb_flow_dissector_target(flow_dissector,
201 FLOW_DISSECTOR_KEY_ENC_KEYID,
202 target_container);
203 keyid->keyid = tunnel_id_to_key32(key->tun_id);
204 }
205
206 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
207 struct flow_dissector_key_ports *tp;
208
209 tp = skb_flow_dissector_target(flow_dissector,
210 FLOW_DISSECTOR_KEY_ENC_PORTS,
211 target_container);
212 tp->src = key->tp_src;
213 tp->dst = key->tp_dst;
214 }
215}
216EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
217
218static enum flow_dissect_ret
219__skb_flow_dissect_mpls(const struct sk_buff *skb,
220 struct flow_dissector *flow_dissector,
221 void *target_container, void *data, int nhoff, int hlen)
222{
223 struct flow_dissector_key_keyid *key_keyid;
224 struct mpls_label *hdr, _hdr[2];
225 u32 entry, label;
226
227 if (!dissector_uses_key(flow_dissector,
228 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
229 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
230 return FLOW_DISSECT_RET_OUT_GOOD;
231
232 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
233 hlen, &_hdr);
234 if (!hdr)
235 return FLOW_DISSECT_RET_OUT_BAD;
236
237 entry = ntohl(hdr[0].entry);
238 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
239
240 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
241 struct flow_dissector_key_mpls *key_mpls;
242
243 key_mpls = skb_flow_dissector_target(flow_dissector,
244 FLOW_DISSECTOR_KEY_MPLS,
245 target_container);
246 key_mpls->mpls_label = label;
247 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
248 >> MPLS_LS_TTL_SHIFT;
249 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
250 >> MPLS_LS_TC_SHIFT;
251 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
252 >> MPLS_LS_S_SHIFT;
253 }
254
255 if (label == MPLS_LABEL_ENTROPY) {
256 key_keyid = skb_flow_dissector_target(flow_dissector,
257 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
258 target_container);
259 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
260 }
261 return FLOW_DISSECT_RET_OUT_GOOD;
262}
263
264static enum flow_dissect_ret
265__skb_flow_dissect_arp(const struct sk_buff *skb,
266 struct flow_dissector *flow_dissector,
267 void *target_container, void *data, int nhoff, int hlen)
268{
269 struct flow_dissector_key_arp *key_arp;
270 struct {
271 unsigned char ar_sha[ETH_ALEN];
272 unsigned char ar_sip[4];
273 unsigned char ar_tha[ETH_ALEN];
274 unsigned char ar_tip[4];
275 } *arp_eth, _arp_eth;
276 const struct arphdr *arp;
277 struct arphdr _arp;
278
279 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
280 return FLOW_DISSECT_RET_OUT_GOOD;
281
282 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
283 hlen, &_arp);
284 if (!arp)
285 return FLOW_DISSECT_RET_OUT_BAD;
286
287 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
288 arp->ar_pro != htons(ETH_P_IP) ||
289 arp->ar_hln != ETH_ALEN ||
290 arp->ar_pln != 4 ||
291 (arp->ar_op != htons(ARPOP_REPLY) &&
292 arp->ar_op != htons(ARPOP_REQUEST)))
293 return FLOW_DISSECT_RET_OUT_BAD;
294
295 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
296 sizeof(_arp_eth), data,
297 hlen, &_arp_eth);
298 if (!arp_eth)
299 return FLOW_DISSECT_RET_OUT_BAD;
300
301 key_arp = skb_flow_dissector_target(flow_dissector,
302 FLOW_DISSECTOR_KEY_ARP,
303 target_container);
304
305 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
306 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
307
308 /* Only store the lower byte of the opcode;
309 * this covers ARPOP_REPLY and ARPOP_REQUEST.
310 */
311 key_arp->op = ntohs(arp->ar_op) & 0xff;
312
313 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
314 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
315
316 return FLOW_DISSECT_RET_OUT_GOOD;
317}
318
319static enum flow_dissect_ret
320__skb_flow_dissect_gre(const struct sk_buff *skb,
321 struct flow_dissector_key_control *key_control,
322 struct flow_dissector *flow_dissector,
323 void *target_container, void *data,
324 __be16 *p_proto, int *p_nhoff, int *p_hlen,
325 unsigned int flags)
326{
327 struct flow_dissector_key_keyid *key_keyid;
328 struct gre_base_hdr *hdr, _hdr;
329 int offset = 0;
330 u16 gre_ver;
331
332 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
333 data, *p_hlen, &_hdr);
334 if (!hdr)
335 return FLOW_DISSECT_RET_OUT_BAD;
336
337 /* Only look inside GRE without routing */
338 if (hdr->flags & GRE_ROUTING)
339 return FLOW_DISSECT_RET_OUT_GOOD;
340
341 /* Only look inside GRE for version 0 and 1 */
342 gre_ver = ntohs(hdr->flags & GRE_VERSION);
343 if (gre_ver > 1)
344 return FLOW_DISSECT_RET_OUT_GOOD;
345
346 *p_proto = hdr->protocol;
347 if (gre_ver) {
348 /* Version1 must be PPTP, and check the flags */
349 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
350 return FLOW_DISSECT_RET_OUT_GOOD;
351 }
352
353 offset += sizeof(struct gre_base_hdr);
354
355 if (hdr->flags & GRE_CSUM)
356 offset += sizeof(((struct gre_full_hdr *) 0)->csum) +
357 sizeof(((struct gre_full_hdr *) 0)->reserved1);
358
359 if (hdr->flags & GRE_KEY) {
360 const __be32 *keyid;
361 __be32 _keyid;
362
363 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
364 sizeof(_keyid),
365 data, *p_hlen, &_keyid);
366 if (!keyid)
367 return FLOW_DISSECT_RET_OUT_BAD;
368
369 if (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,
373 target_container);
374 if (gre_ver == 0)
375 key_keyid->keyid = *keyid;
376 else
377 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
378 }
379 offset += sizeof(((struct gre_full_hdr *) 0)->key);
380 }
381
382 if (hdr->flags & GRE_SEQ)
383 offset += sizeof(((struct pptp_gre_header *) 0)->seq);
384
385 if (gre_ver == 0) {
386 if (*p_proto == htons(ETH_P_TEB)) {
387 const struct ethhdr *eth;
388 struct ethhdr _eth;
389
390 eth = __skb_header_pointer(skb, *p_nhoff + offset,
391 sizeof(_eth),
392 data, *p_hlen, &_eth);
393 if (!eth)
394 return FLOW_DISSECT_RET_OUT_BAD;
395 *p_proto = eth->h_proto;
396 offset += sizeof(*eth);
397
398 /* Cap headers that we access via pointers at the
399 * end of the Ethernet header as our maximum alignment
400 * at that point is only 2 bytes.
401 */
402 if (NET_IP_ALIGN)
403 *p_hlen = *p_nhoff + offset;
404 }
405 } else { /* version 1, must be PPTP */
406 u8 _ppp_hdr[PPP_HDRLEN];
407 u8 *ppp_hdr;
408
409 if (hdr->flags & GRE_ACK)
410 offset += sizeof(((struct pptp_gre_header *) 0)->ack);
411
412 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
413 sizeof(_ppp_hdr),
414 data, *p_hlen, _ppp_hdr);
415 if (!ppp_hdr)
416 return FLOW_DISSECT_RET_OUT_BAD;
417
418 switch (PPP_PROTOCOL(ppp_hdr)) {
419 case PPP_IP:
420 *p_proto = htons(ETH_P_IP);
421 break;
422 case PPP_IPV6:
423 *p_proto = htons(ETH_P_IPV6);
424 break;
425 default:
426 /* Could probably catch some more like MPLS */
427 break;
428 }
429
430 offset += PPP_HDRLEN;
431 }
432
433 *p_nhoff += offset;
434 key_control->flags |= FLOW_DIS_ENCAPSULATION;
435 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
436 return FLOW_DISSECT_RET_OUT_GOOD;
437
438 return FLOW_DISSECT_RET_PROTO_AGAIN;
439}
440
441/**
442 * __skb_flow_dissect_batadv() - dissect batman-adv header
443 * @skb: sk_buff to with the batman-adv header
444 * @key_control: flow dissectors control key
445 * @data: raw buffer pointer to the packet, if NULL use skb->data
446 * @p_proto: pointer used to update the protocol to process next
447 * @p_nhoff: pointer used to update inner network header offset
448 * @hlen: packet header length
449 * @flags: any combination of FLOW_DISSECTOR_F_*
450 *
451 * ETH_P_BATMAN packets are tried to be dissected. Only
452 * &struct batadv_unicast packets are actually processed because they contain an
453 * inner ethernet header and are usually followed by actual network header. This
454 * allows the flow dissector to continue processing the packet.
455 *
456 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
457 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
458 * otherwise FLOW_DISSECT_RET_OUT_BAD
459 */
460static enum flow_dissect_ret
461__skb_flow_dissect_batadv(const struct sk_buff *skb,
462 struct flow_dissector_key_control *key_control,
463 void *data, __be16 *p_proto, int *p_nhoff, int hlen,
464 unsigned int flags)
465{
466 struct {
467 struct batadv_unicast_packet batadv_unicast;
468 struct ethhdr eth;
469 } *hdr, _hdr;
470
471 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
472 &_hdr);
473 if (!hdr)
474 return FLOW_DISSECT_RET_OUT_BAD;
475
476 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
477 return FLOW_DISSECT_RET_OUT_BAD;
478
479 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
480 return FLOW_DISSECT_RET_OUT_BAD;
481
482 *p_proto = hdr->eth.h_proto;
483 *p_nhoff += sizeof(*hdr);
484
485 key_control->flags |= FLOW_DIS_ENCAPSULATION;
486 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
487 return FLOW_DISSECT_RET_OUT_GOOD;
488
489 return FLOW_DISSECT_RET_PROTO_AGAIN;
490}
491
492static void
493__skb_flow_dissect_tcp(const struct sk_buff *skb,
494 struct flow_dissector *flow_dissector,
495 void *target_container, void *data, int thoff, int hlen)
496{
497 struct flow_dissector_key_tcp *key_tcp;
498 struct tcphdr *th, _th;
499
500 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
501 return;
502
503 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
504 if (!th)
505 return;
506
507 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
508 return;
509
510 key_tcp = skb_flow_dissector_target(flow_dissector,
511 FLOW_DISSECTOR_KEY_TCP,
512 target_container);
513 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
514}
515
516static void
517__skb_flow_dissect_ipv4(const struct sk_buff *skb,
518 struct flow_dissector *flow_dissector,
519 void *target_container, void *data, const struct iphdr *iph)
520{
521 struct flow_dissector_key_ip *key_ip;
522
523 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
524 return;
525
526 key_ip = skb_flow_dissector_target(flow_dissector,
527 FLOW_DISSECTOR_KEY_IP,
528 target_container);
529 key_ip->tos = iph->tos;
530 key_ip->ttl = iph->ttl;
531}
532
533static void
534__skb_flow_dissect_ipv6(const struct sk_buff *skb,
535 struct flow_dissector *flow_dissector,
536 void *target_container, void *data, const struct ipv6hdr *iph)
537{
538 struct flow_dissector_key_ip *key_ip;
539
540 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
541 return;
542
543 key_ip = skb_flow_dissector_target(flow_dissector,
544 FLOW_DISSECTOR_KEY_IP,
545 target_container);
546 key_ip->tos = ipv6_get_dsfield(iph);
547 key_ip->ttl = iph->hop_limit;
548}
549
550/* Maximum number of protocol headers that can be parsed in
551 * __skb_flow_dissect
552 */
553#define MAX_FLOW_DISSECT_HDRS 15
554
555static bool skb_flow_dissect_allowed(int *num_hdrs)
556{
557 ++*num_hdrs;
558
559 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
560}
561
562/**
563 * __skb_flow_dissect - extract the flow_keys struct and return it
564 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
565 * @flow_dissector: list of keys to dissect
566 * @target_container: target structure to put dissected values into
567 * @data: raw buffer pointer to the packet, if NULL use skb->data
568 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
569 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
570 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
571 *
572 * The function will try to retrieve individual keys into target specified
573 * by flow_dissector from either the skbuff or a raw buffer specified by the
574 * rest parameters.
575 *
576 * Caller must take care of zeroing target container memory.
577 */
578bool __skb_flow_dissect(const struct sk_buff *skb,
579 struct flow_dissector *flow_dissector,
580 void *target_container,
581 void *data, __be16 proto, int nhoff, int hlen,
582 unsigned int flags)
583{
584 struct flow_dissector_key_control *key_control;
585 struct flow_dissector_key_basic *key_basic;
586 struct flow_dissector_key_addrs *key_addrs;
587 struct flow_dissector_key_ports *key_ports;
588 struct flow_dissector_key_icmp *key_icmp;
589 struct flow_dissector_key_tags *key_tags;
590 struct flow_dissector_key_vlan *key_vlan;
591 enum flow_dissect_ret fdret;
592 bool skip_vlan = false;
593 int num_hdrs = 0;
594 u8 ip_proto = 0;
595 bool ret;
596
597 if (!data) {
598 data = skb->data;
599 proto = skb_vlan_tag_present(skb) ?
600 skb->vlan_proto : skb->protocol;
601 nhoff = skb_network_offset(skb);
602 hlen = skb_headlen(skb);
603#if IS_ENABLED(CONFIG_NET_DSA)
604 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
605 const struct dsa_device_ops *ops;
606 int offset;
607
608 ops = skb->dev->dsa_ptr->tag_ops;
609 if (ops->flow_dissect &&
610 !ops->flow_dissect(skb, &proto, &offset)) {
611 hlen -= offset;
612 nhoff += offset;
613 }
614 }
615#endif
616 }
617
618 /* It is ensured by skb_flow_dissector_init() that control key will
619 * be always present.
620 */
621 key_control = skb_flow_dissector_target(flow_dissector,
622 FLOW_DISSECTOR_KEY_CONTROL,
623 target_container);
624
625 /* It is ensured by skb_flow_dissector_init() that basic key will
626 * be always present.
627 */
628 key_basic = skb_flow_dissector_target(flow_dissector,
629 FLOW_DISSECTOR_KEY_BASIC,
630 target_container);
631
632 if (dissector_uses_key(flow_dissector,
633 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
634 struct ethhdr *eth = eth_hdr(skb);
635 struct flow_dissector_key_eth_addrs *key_eth_addrs;
636
637 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
638 FLOW_DISSECTOR_KEY_ETH_ADDRS,
639 target_container);
640 memcpy(key_eth_addrs, ð->h_dest, sizeof(*key_eth_addrs));
641 }
642
643proto_again:
644 fdret = FLOW_DISSECT_RET_CONTINUE;
645
646 switch (proto) {
647 case htons(ETH_P_IP): {
648 const struct iphdr *iph;
649 struct iphdr _iph;
650
651 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
652 if (!iph || iph->ihl < 5) {
653 fdret = FLOW_DISSECT_RET_OUT_BAD;
654 break;
655 }
656
657 nhoff += iph->ihl * 4;
658
659 ip_proto = iph->protocol;
660
661 if (dissector_uses_key(flow_dissector,
662 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
663 key_addrs = skb_flow_dissector_target(flow_dissector,
664 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
665 target_container);
666
667 memcpy(&key_addrs->v4addrs, &iph->saddr,
668 sizeof(key_addrs->v4addrs));
669 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
670 }
671
672 if (ip_is_fragment(iph)) {
673 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
674
675 if (iph->frag_off & htons(IP_OFFSET)) {
676 fdret = FLOW_DISSECT_RET_OUT_GOOD;
677 break;
678 } else {
679 key_control->flags |= FLOW_DIS_FIRST_FRAG;
680 if (!(flags &
681 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
682 fdret = FLOW_DISSECT_RET_OUT_GOOD;
683 break;
684 }
685 }
686 }
687
688 __skb_flow_dissect_ipv4(skb, flow_dissector,
689 target_container, data, iph);
690
691 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
692 fdret = FLOW_DISSECT_RET_OUT_GOOD;
693 break;
694 }
695
696 break;
697 }
698 case htons(ETH_P_IPV6): {
699 const struct ipv6hdr *iph;
700 struct ipv6hdr _iph;
701
702 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
703 if (!iph) {
704 fdret = FLOW_DISSECT_RET_OUT_BAD;
705 break;
706 }
707
708 ip_proto = iph->nexthdr;
709 nhoff += sizeof(struct ipv6hdr);
710
711 if (dissector_uses_key(flow_dissector,
712 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
713 key_addrs = skb_flow_dissector_target(flow_dissector,
714 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
715 target_container);
716
717 memcpy(&key_addrs->v6addrs, &iph->saddr,
718 sizeof(key_addrs->v6addrs));
719 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
720 }
721
722 if ((dissector_uses_key(flow_dissector,
723 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
724 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
725 ip6_flowlabel(iph)) {
726 __be32 flow_label = ip6_flowlabel(iph);
727
728 if (dissector_uses_key(flow_dissector,
729 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
730 key_tags = skb_flow_dissector_target(flow_dissector,
731 FLOW_DISSECTOR_KEY_FLOW_LABEL,
732 target_container);
733 key_tags->flow_label = ntohl(flow_label);
734 }
735 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
736 fdret = FLOW_DISSECT_RET_OUT_GOOD;
737 break;
738 }
739 }
740
741 __skb_flow_dissect_ipv6(skb, flow_dissector,
742 target_container, data, iph);
743
744 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
745 fdret = FLOW_DISSECT_RET_OUT_GOOD;
746
747 break;
748 }
749 case htons(ETH_P_8021AD):
750 case htons(ETH_P_8021Q): {
751 const struct vlan_hdr *vlan;
752 struct vlan_hdr _vlan;
753 bool vlan_tag_present = skb && skb_vlan_tag_present(skb);
754
755 if (vlan_tag_present)
756 proto = skb->protocol;
757
758 if (!vlan_tag_present || eth_type_vlan(skb->protocol)) {
759 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
760 data, hlen, &_vlan);
761 if (!vlan) {
762 fdret = FLOW_DISSECT_RET_OUT_BAD;
763 break;
764 }
765
766 proto = vlan->h_vlan_encapsulated_proto;
767 nhoff += sizeof(*vlan);
768 if (skip_vlan) {
769 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
770 break;
771 }
772 }
773
774 skip_vlan = true;
775 if (dissector_uses_key(flow_dissector,
776 FLOW_DISSECTOR_KEY_VLAN)) {
777 key_vlan = skb_flow_dissector_target(flow_dissector,
778 FLOW_DISSECTOR_KEY_VLAN,
779 target_container);
780
781 if (vlan_tag_present) {
782 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
783 key_vlan->vlan_priority =
784 (skb_vlan_tag_get_prio(skb) >> VLAN_PRIO_SHIFT);
785 } else {
786 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
787 VLAN_VID_MASK;
788 key_vlan->vlan_priority =
789 (ntohs(vlan->h_vlan_TCI) &
790 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
791 }
792 }
793
794 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
795 break;
796 }
797 case htons(ETH_P_PPP_SES): {
798 struct {
799 struct pppoe_hdr hdr;
800 __be16 proto;
801 } *hdr, _hdr;
802 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
803 if (!hdr) {
804 fdret = FLOW_DISSECT_RET_OUT_BAD;
805 break;
806 }
807
808 proto = hdr->proto;
809 nhoff += PPPOE_SES_HLEN;
810 switch (proto) {
811 case htons(PPP_IP):
812 proto = htons(ETH_P_IP);
813 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
814 break;
815 case htons(PPP_IPV6):
816 proto = htons(ETH_P_IPV6);
817 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
818 break;
819 default:
820 fdret = FLOW_DISSECT_RET_OUT_BAD;
821 break;
822 }
823 break;
824 }
825 case htons(ETH_P_TIPC): {
826 struct tipc_basic_hdr *hdr, _hdr;
827
828 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
829 data, hlen, &_hdr);
830 if (!hdr) {
831 fdret = FLOW_DISSECT_RET_OUT_BAD;
832 break;
833 }
834
835 if (dissector_uses_key(flow_dissector,
836 FLOW_DISSECTOR_KEY_TIPC)) {
837 key_addrs = skb_flow_dissector_target(flow_dissector,
838 FLOW_DISSECTOR_KEY_TIPC,
839 target_container);
840 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
841 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
842 }
843 fdret = FLOW_DISSECT_RET_OUT_GOOD;
844 break;
845 }
846
847 case htons(ETH_P_MPLS_UC):
848 case htons(ETH_P_MPLS_MC):
849 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
850 target_container, data,
851 nhoff, hlen);
852 break;
853 case htons(ETH_P_FCOE):
854 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
855 fdret = FLOW_DISSECT_RET_OUT_BAD;
856 break;
857 }
858
859 nhoff += FCOE_HEADER_LEN;
860 fdret = FLOW_DISSECT_RET_OUT_GOOD;
861 break;
862
863 case htons(ETH_P_ARP):
864 case htons(ETH_P_RARP):
865 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
866 target_container, data,
867 nhoff, hlen);
868 break;
869
870 case htons(ETH_P_BATMAN):
871 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
872 &proto, &nhoff, hlen, flags);
873 break;
874
875 default:
876 fdret = FLOW_DISSECT_RET_OUT_BAD;
877 break;
878 }
879
880 /* Process result of proto processing */
881 switch (fdret) {
882 case FLOW_DISSECT_RET_OUT_GOOD:
883 goto out_good;
884 case FLOW_DISSECT_RET_PROTO_AGAIN:
885 if (skb_flow_dissect_allowed(&num_hdrs))
886 goto proto_again;
887 goto out_good;
888 case FLOW_DISSECT_RET_CONTINUE:
889 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
890 break;
891 case FLOW_DISSECT_RET_OUT_BAD:
892 default:
893 goto out_bad;
894 }
895
896ip_proto_again:
897 fdret = FLOW_DISSECT_RET_CONTINUE;
898
899 switch (ip_proto) {
900 case IPPROTO_GRE:
901 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
902 target_container, data,
903 &proto, &nhoff, &hlen, flags);
904 break;
905
906 case NEXTHDR_HOP:
907 case NEXTHDR_ROUTING:
908 case NEXTHDR_DEST: {
909 u8 _opthdr[2], *opthdr;
910
911 if (proto != htons(ETH_P_IPV6))
912 break;
913
914 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
915 data, hlen, &_opthdr);
916 if (!opthdr) {
917 fdret = FLOW_DISSECT_RET_OUT_BAD;
918 break;
919 }
920
921 ip_proto = opthdr[0];
922 nhoff += (opthdr[1] + 1) << 3;
923
924 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
925 break;
926 }
927 case NEXTHDR_FRAGMENT: {
928 struct frag_hdr _fh, *fh;
929
930 if (proto != htons(ETH_P_IPV6))
931 break;
932
933 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
934 data, hlen, &_fh);
935
936 if (!fh) {
937 fdret = FLOW_DISSECT_RET_OUT_BAD;
938 break;
939 }
940
941 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
942
943 nhoff += sizeof(_fh);
944 ip_proto = fh->nexthdr;
945
946 if (!(fh->frag_off & htons(IP6_OFFSET))) {
947 key_control->flags |= FLOW_DIS_FIRST_FRAG;
948 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
949 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
950 break;
951 }
952 }
953
954 fdret = FLOW_DISSECT_RET_OUT_GOOD;
955 break;
956 }
957 case IPPROTO_IPIP:
958 proto = htons(ETH_P_IP);
959
960 key_control->flags |= FLOW_DIS_ENCAPSULATION;
961 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
962 fdret = FLOW_DISSECT_RET_OUT_GOOD;
963 break;
964 }
965
966 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
967 break;
968
969 case IPPROTO_IPV6:
970 proto = htons(ETH_P_IPV6);
971
972 key_control->flags |= FLOW_DIS_ENCAPSULATION;
973 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
974 fdret = FLOW_DISSECT_RET_OUT_GOOD;
975 break;
976 }
977
978 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
979 break;
980
981
982 case IPPROTO_MPLS:
983 proto = htons(ETH_P_MPLS_UC);
984 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
985 break;
986
987 case IPPROTO_TCP:
988 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
989 data, nhoff, hlen);
990 break;
991
992 default:
993 break;
994 }
995
996 if (dissector_uses_key(flow_dissector,
997 FLOW_DISSECTOR_KEY_PORTS)) {
998 key_ports = skb_flow_dissector_target(flow_dissector,
999 FLOW_DISSECTOR_KEY_PORTS,
1000 target_container);
1001 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1002 data, hlen);
1003 }
1004
1005 if (dissector_uses_key(flow_dissector,
1006 FLOW_DISSECTOR_KEY_ICMP)) {
1007 key_icmp = skb_flow_dissector_target(flow_dissector,
1008 FLOW_DISSECTOR_KEY_ICMP,
1009 target_container);
1010 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1011 }
1012
1013 /* Process result of IP proto processing */
1014 switch (fdret) {
1015 case FLOW_DISSECT_RET_PROTO_AGAIN:
1016 if (skb_flow_dissect_allowed(&num_hdrs))
1017 goto proto_again;
1018 break;
1019 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1020 if (skb_flow_dissect_allowed(&num_hdrs))
1021 goto ip_proto_again;
1022 break;
1023 case FLOW_DISSECT_RET_OUT_GOOD:
1024 case FLOW_DISSECT_RET_CONTINUE:
1025 break;
1026 case FLOW_DISSECT_RET_OUT_BAD:
1027 default:
1028 goto out_bad;
1029 }
1030
1031out_good:
1032 ret = true;
1033
1034out:
1035 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1036 key_basic->n_proto = proto;
1037 key_basic->ip_proto = ip_proto;
1038
1039 return ret;
1040
1041out_bad:
1042 ret = false;
1043 goto out;
1044}
1045EXPORT_SYMBOL(__skb_flow_dissect);
1046
1047static u32 hashrnd __read_mostly;
1048static __always_inline void __flow_hash_secret_init(void)
1049{
1050 net_get_random_once(&hashrnd, sizeof(hashrnd));
1051}
1052
1053static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1054 u32 keyval)
1055{
1056 return jhash2(words, length, keyval);
1057}
1058
1059static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1060{
1061 const void *p = flow;
1062
1063 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1064 return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1065}
1066
1067static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1068{
1069 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1070 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1071 BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1072 sizeof(*flow) - sizeof(flow->addrs));
1073
1074 switch (flow->control.addr_type) {
1075 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1076 diff -= sizeof(flow->addrs.v4addrs);
1077 break;
1078 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1079 diff -= sizeof(flow->addrs.v6addrs);
1080 break;
1081 case FLOW_DISSECTOR_KEY_TIPC:
1082 diff -= sizeof(flow->addrs.tipckey);
1083 break;
1084 }
1085 return (sizeof(*flow) - diff) / sizeof(u32);
1086}
1087
1088__be32 flow_get_u32_src(const struct flow_keys *flow)
1089{
1090 switch (flow->control.addr_type) {
1091 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1092 return flow->addrs.v4addrs.src;
1093 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1094 return (__force __be32)ipv6_addr_hash(
1095 &flow->addrs.v6addrs.src);
1096 case FLOW_DISSECTOR_KEY_TIPC:
1097 return flow->addrs.tipckey.key;
1098 default:
1099 return 0;
1100 }
1101}
1102EXPORT_SYMBOL(flow_get_u32_src);
1103
1104__be32 flow_get_u32_dst(const struct flow_keys *flow)
1105{
1106 switch (flow->control.addr_type) {
1107 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1108 return flow->addrs.v4addrs.dst;
1109 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1110 return (__force __be32)ipv6_addr_hash(
1111 &flow->addrs.v6addrs.dst);
1112 default:
1113 return 0;
1114 }
1115}
1116EXPORT_SYMBOL(flow_get_u32_dst);
1117
1118static inline void __flow_hash_consistentify(struct flow_keys *keys)
1119{
1120 int addr_diff, i;
1121
1122 switch (keys->control.addr_type) {
1123 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1124 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1125 (__force u32)keys->addrs.v4addrs.src;
1126 if ((addr_diff < 0) ||
1127 (addr_diff == 0 &&
1128 ((__force u16)keys->ports.dst <
1129 (__force u16)keys->ports.src))) {
1130 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1131 swap(keys->ports.src, keys->ports.dst);
1132 }
1133 break;
1134 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1135 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1136 &keys->addrs.v6addrs.src,
1137 sizeof(keys->addrs.v6addrs.dst));
1138 if ((addr_diff < 0) ||
1139 (addr_diff == 0 &&
1140 ((__force u16)keys->ports.dst <
1141 (__force u16)keys->ports.src))) {
1142 for (i = 0; i < 4; i++)
1143 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1144 keys->addrs.v6addrs.dst.s6_addr32[i]);
1145 swap(keys->ports.src, keys->ports.dst);
1146 }
1147 break;
1148 }
1149}
1150
1151static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1152{
1153 u32 hash;
1154
1155 __flow_hash_consistentify(keys);
1156
1157 hash = __flow_hash_words(flow_keys_hash_start(keys),
1158 flow_keys_hash_length(keys), keyval);
1159 if (!hash)
1160 hash = 1;
1161
1162 return hash;
1163}
1164
1165u32 flow_hash_from_keys(struct flow_keys *keys)
1166{
1167 __flow_hash_secret_init();
1168 return __flow_hash_from_keys(keys, hashrnd);
1169}
1170EXPORT_SYMBOL(flow_hash_from_keys);
1171
1172static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1173 struct flow_keys *keys, u32 keyval)
1174{
1175 skb_flow_dissect_flow_keys(skb, keys,
1176 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1177
1178 return __flow_hash_from_keys(keys, keyval);
1179}
1180
1181struct _flow_keys_digest_data {
1182 __be16 n_proto;
1183 u8 ip_proto;
1184 u8 padding;
1185 __be32 ports;
1186 __be32 src;
1187 __be32 dst;
1188};
1189
1190void make_flow_keys_digest(struct flow_keys_digest *digest,
1191 const struct flow_keys *flow)
1192{
1193 struct _flow_keys_digest_data *data =
1194 (struct _flow_keys_digest_data *)digest;
1195
1196 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1197
1198 memset(digest, 0, sizeof(*digest));
1199
1200 data->n_proto = flow->basic.n_proto;
1201 data->ip_proto = flow->basic.ip_proto;
1202 data->ports = flow->ports.ports;
1203 data->src = flow->addrs.v4addrs.src;
1204 data->dst = flow->addrs.v4addrs.dst;
1205}
1206EXPORT_SYMBOL(make_flow_keys_digest);
1207
1208static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1209
1210u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1211{
1212 struct flow_keys keys;
1213
1214 __flow_hash_secret_init();
1215
1216 memset(&keys, 0, sizeof(keys));
1217 __skb_flow_dissect(skb, &flow_keys_dissector_symmetric, &keys,
1218 NULL, 0, 0, 0,
1219 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1220
1221 return __flow_hash_from_keys(&keys, hashrnd);
1222}
1223EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1224
1225/**
1226 * __skb_get_hash: calculate a flow hash
1227 * @skb: sk_buff to calculate flow hash from
1228 *
1229 * This function calculates a flow hash based on src/dst addresses
1230 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1231 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1232 * if hash is a canonical 4-tuple hash over transport ports.
1233 */
1234void __skb_get_hash(struct sk_buff *skb)
1235{
1236 struct flow_keys keys;
1237 u32 hash;
1238
1239 __flow_hash_secret_init();
1240
1241 hash = ___skb_get_hash(skb, &keys, hashrnd);
1242
1243 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1244}
1245EXPORT_SYMBOL(__skb_get_hash);
1246
1247__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1248{
1249 struct flow_keys keys;
1250
1251 return ___skb_get_hash(skb, &keys, perturb);
1252}
1253EXPORT_SYMBOL(skb_get_hash_perturb);
1254
1255u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1256 const struct flow_keys *keys, int hlen)
1257{
1258 u32 poff = keys->control.thoff;
1259
1260 /* skip L4 headers for fragments after the first */
1261 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1262 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1263 return poff;
1264
1265 switch (keys->basic.ip_proto) {
1266 case IPPROTO_TCP: {
1267 /* access doff as u8 to avoid unaligned access */
1268 const u8 *doff;
1269 u8 _doff;
1270
1271 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1272 data, hlen, &_doff);
1273 if (!doff)
1274 return poff;
1275
1276 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1277 break;
1278 }
1279 case IPPROTO_UDP:
1280 case IPPROTO_UDPLITE:
1281 poff += sizeof(struct udphdr);
1282 break;
1283 /* For the rest, we do not really care about header
1284 * extensions at this point for now.
1285 */
1286 case IPPROTO_ICMP:
1287 poff += sizeof(struct icmphdr);
1288 break;
1289 case IPPROTO_ICMPV6:
1290 poff += sizeof(struct icmp6hdr);
1291 break;
1292 case IPPROTO_IGMP:
1293 poff += sizeof(struct igmphdr);
1294 break;
1295 case IPPROTO_DCCP:
1296 poff += sizeof(struct dccp_hdr);
1297 break;
1298 case IPPROTO_SCTP:
1299 poff += sizeof(struct sctphdr);
1300 break;
1301 }
1302
1303 return poff;
1304}
1305
1306/**
1307 * skb_get_poff - get the offset to the payload
1308 * @skb: sk_buff to get the payload offset from
1309 *
1310 * The function will get the offset to the payload as far as it could
1311 * be dissected. The main user is currently BPF, so that we can dynamically
1312 * truncate packets without needing to push actual payload to the user
1313 * space and can analyze headers only, instead.
1314 */
1315u32 skb_get_poff(const struct sk_buff *skb)
1316{
1317 struct flow_keys keys;
1318
1319 if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
1320 return 0;
1321
1322 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1323}
1324
1325__u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1326{
1327 memset(keys, 0, sizeof(*keys));
1328
1329 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1330 sizeof(keys->addrs.v6addrs.src));
1331 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1332 sizeof(keys->addrs.v6addrs.dst));
1333 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1334 keys->ports.src = fl6->fl6_sport;
1335 keys->ports.dst = fl6->fl6_dport;
1336 keys->keyid.keyid = fl6->fl6_gre_key;
1337 keys->tags.flow_label = (__force u32)fl6->flowlabel;
1338 keys->basic.ip_proto = fl6->flowi6_proto;
1339
1340 return flow_hash_from_keys(keys);
1341}
1342EXPORT_SYMBOL(__get_hash_from_flowi6);
1343
1344static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1345 {
1346 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1347 .offset = offsetof(struct flow_keys, control),
1348 },
1349 {
1350 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1351 .offset = offsetof(struct flow_keys, basic),
1352 },
1353 {
1354 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1355 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1356 },
1357 {
1358 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1359 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1360 },
1361 {
1362 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1363 .offset = offsetof(struct flow_keys, addrs.tipckey),
1364 },
1365 {
1366 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1367 .offset = offsetof(struct flow_keys, ports),
1368 },
1369 {
1370 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1371 .offset = offsetof(struct flow_keys, vlan),
1372 },
1373 {
1374 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1375 .offset = offsetof(struct flow_keys, tags),
1376 },
1377 {
1378 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1379 .offset = offsetof(struct flow_keys, keyid),
1380 },
1381};
1382
1383static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1384 {
1385 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1386 .offset = offsetof(struct flow_keys, control),
1387 },
1388 {
1389 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1390 .offset = offsetof(struct flow_keys, basic),
1391 },
1392 {
1393 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1394 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1395 },
1396 {
1397 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1398 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1399 },
1400 {
1401 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1402 .offset = offsetof(struct flow_keys, ports),
1403 },
1404};
1405
1406static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
1407 {
1408 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1409 .offset = offsetof(struct flow_keys, control),
1410 },
1411 {
1412 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1413 .offset = offsetof(struct flow_keys, basic),
1414 },
1415};
1416
1417struct flow_dissector flow_keys_dissector __read_mostly;
1418EXPORT_SYMBOL(flow_keys_dissector);
1419
1420struct flow_dissector flow_keys_buf_dissector __read_mostly;
1421
1422static int __init init_default_flow_dissectors(void)
1423{
1424 skb_flow_dissector_init(&flow_keys_dissector,
1425 flow_keys_dissector_keys,
1426 ARRAY_SIZE(flow_keys_dissector_keys));
1427 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1428 flow_keys_dissector_symmetric_keys,
1429 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1430 skb_flow_dissector_init(&flow_keys_buf_dissector,
1431 flow_keys_buf_dissector_keys,
1432 ARRAY_SIZE(flow_keys_buf_dissector_keys));
1433 return 0;
1434}
1435
1436core_initcall(init_default_flow_dissectors);
1// SPDX-License-Identifier: GPL-2.0-only
2#include <linux/kernel.h>
3#include <linux/skbuff.h>
4#include <linux/export.h>
5#include <linux/ip.h>
6#include <linux/ipv6.h>
7#include <linux/if_vlan.h>
8#include <linux/filter.h>
9#include <net/dsa.h>
10#include <net/dst_metadata.h>
11#include <net/ip.h>
12#include <net/ipv6.h>
13#include <net/gre.h>
14#include <net/pptp.h>
15#include <net/tipc.h>
16#include <linux/igmp.h>
17#include <linux/icmp.h>
18#include <linux/sctp.h>
19#include <linux/dccp.h>
20#include <linux/if_tunnel.h>
21#include <linux/if_pppox.h>
22#include <linux/ppp_defs.h>
23#include <linux/stddef.h>
24#include <linux/if_ether.h>
25#include <linux/if_hsr.h>
26#include <linux/mpls.h>
27#include <linux/tcp.h>
28#include <linux/ptp_classify.h>
29#include <net/flow_dissector.h>
30#include <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>
36#endif
37#include <linux/bpf-netns.h>
38
39static void dissector_set_key(struct flow_dissector *flow_dissector,
40 enum flow_dissector_key_id key_id)
41{
42 flow_dissector->used_keys |= (1 << key_id);
43}
44
45void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
46 const struct flow_dissector_key *key,
47 unsigned int key_count)
48{
49 unsigned int i;
50
51 memset(flow_dissector, 0, sizeof(*flow_dissector));
52
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.
56 */
57 BUG_ON(key->offset > USHRT_MAX);
58 BUG_ON(dissector_uses_key(flow_dissector,
59 key->key_id));
60
61 dissector_set_key(flow_dissector, key->key_id);
62 flow_dissector->offset[key->key_id] = key->offset;
63 }
64
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.
67 */
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));
72}
73EXPORT_SYMBOL(skb_flow_dissector_init);
74
75#ifdef CONFIG_BPF_SYSCALL
76int flow_dissector_bpf_prog_attach_check(struct net *net,
77 struct bpf_prog *prog)
78{
79 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
80
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.
86 */
87 struct net *ns;
88
89 for_each_net(ns) {
90 if (ns == &init_net)
91 continue;
92 if (rcu_access_pointer(ns->bpf.run_array[type]))
93 return -EEXIST;
94 }
95 } else {
96 /* Make sure root flow dissector is not attached
97 * when attaching to the non-root namespace.
98 */
99 if (rcu_access_pointer(init_net.bpf.run_array[type]))
100 return -EEXIST;
101 }
102
103 return 0;
104}
105#endif /* CONFIG_BPF_SYSCALL */
106
107/**
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)
114 *
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
117 */
118__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
119 const void *data, int hlen)
120{
121 int poff = proto_ports_offset(ip_proto);
122
123 if (!data) {
124 data = skb->data;
125 hlen = skb_headlen(skb);
126 }
127
128 if (poff >= 0) {
129 __be32 *ports, _ports;
130
131 ports = __skb_header_pointer(skb, thoff + poff,
132 sizeof(_ports), data, hlen, &_ports);
133 if (ports)
134 return *ports;
135 }
136
137 return 0;
138}
139EXPORT_SYMBOL(__skb_flow_get_ports);
140
141static bool icmp_has_id(u8 type)
142{
143 switch (type) {
144 case ICMP_ECHO:
145 case ICMP_ECHOREPLY:
146 case ICMP_TIMESTAMP:
147 case ICMP_TIMESTAMPREPLY:
148 case ICMPV6_ECHO_REQUEST:
149 case ICMPV6_ECHO_REPLY:
150 return true;
151 }
152
153 return false;
154}
155
156/**
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
163 */
164void 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)
167{
168 struct icmphdr *ih, _ih;
169
170 ih = __skb_header_pointer(skb, thoff, sizeof(_ih), data, hlen, &_ih);
171 if (!ih)
172 return;
173
174 key_icmp->type = ih->type;
175 key_icmp->code = ih->code;
176
177 /* As we use 0 to signal that the Id field is not present,
178 * avoid confusion with packets without such field
179 */
180 if (icmp_has_id(ih->type))
181 key_icmp->id = ih->un.echo.id ? ntohs(ih->un.echo.id) : 1;
182 else
183 key_icmp->id = 0;
184}
185EXPORT_SYMBOL(skb_flow_get_icmp_tci);
186
187/* If FLOW_DISSECTOR_KEY_ICMP is set, dissect an ICMP packet
188 * using skb_flow_get_icmp_tci().
189 */
190static void __skb_flow_dissect_icmp(const struct sk_buff *skb,
191 struct flow_dissector *flow_dissector,
192 void *target_container, const void *data,
193 int thoff, int hlen)
194{
195 struct flow_dissector_key_icmp *key_icmp;
196
197 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ICMP))
198 return;
199
200 key_icmp = skb_flow_dissector_target(flow_dissector,
201 FLOW_DISSECTOR_KEY_ICMP,
202 target_container);
203
204 skb_flow_get_icmp_tci(skb, key_icmp, data, thoff, hlen);
205}
206
207static void __skb_flow_dissect_l2tpv3(const struct sk_buff *skb,
208 struct flow_dissector *flow_dissector,
209 void *target_container, const void *data,
210 int nhoff, int hlen)
211{
212 struct flow_dissector_key_l2tpv3 *key_l2tpv3;
213 struct {
214 __be32 session_id;
215 } *hdr, _hdr;
216
217 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_L2TPV3))
218 return;
219
220 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
221 if (!hdr)
222 return;
223
224 key_l2tpv3 = skb_flow_dissector_target(flow_dissector,
225 FLOW_DISSECTOR_KEY_L2TPV3,
226 target_container);
227
228 key_l2tpv3->session_id = hdr->session_id;
229}
230
231void skb_flow_dissect_meta(const struct sk_buff *skb,
232 struct flow_dissector *flow_dissector,
233 void *target_container)
234{
235 struct flow_dissector_key_meta *meta;
236
237 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_META))
238 return;
239
240 meta = skb_flow_dissector_target(flow_dissector,
241 FLOW_DISSECTOR_KEY_META,
242 target_container);
243 meta->ingress_ifindex = skb->skb_iif;
244}
245EXPORT_SYMBOL(skb_flow_dissect_meta);
246
247static void
248skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
249 struct flow_dissector *flow_dissector,
250 void *target_container)
251{
252 struct flow_dissector_key_control *ctrl;
253
254 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
255 return;
256
257 ctrl = skb_flow_dissector_target(flow_dissector,
258 FLOW_DISSECTOR_KEY_ENC_CONTROL,
259 target_container);
260 ctrl->addr_type = type;
261}
262
263void
264skb_flow_dissect_ct(const struct sk_buff *skb,
265 struct flow_dissector *flow_dissector,
266 void *target_container, u16 *ctinfo_map,
267 size_t mapsize, bool post_ct, u16 zone)
268{
269#if IS_ENABLED(CONFIG_NF_CONNTRACK)
270 struct flow_dissector_key_ct *key;
271 enum ip_conntrack_info ctinfo;
272 struct nf_conn_labels *cl;
273 struct nf_conn *ct;
274
275 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_CT))
276 return;
277
278 ct = nf_ct_get(skb, &ctinfo);
279 if (!ct && !post_ct)
280 return;
281
282 key = skb_flow_dissector_target(flow_dissector,
283 FLOW_DISSECTOR_KEY_CT,
284 target_container);
285
286 if (!ct) {
287 key->ct_state = TCA_FLOWER_KEY_CT_FLAGS_TRACKED |
288 TCA_FLOWER_KEY_CT_FLAGS_INVALID;
289 key->ct_zone = zone;
290 return;
291 }
292
293 if (ctinfo < mapsize)
294 key->ct_state = ctinfo_map[ctinfo];
295#if IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)
296 key->ct_zone = ct->zone.id;
297#endif
298#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
299 key->ct_mark = READ_ONCE(ct->mark);
300#endif
301
302 cl = nf_ct_labels_find(ct);
303 if (cl)
304 memcpy(key->ct_labels, cl->bits, sizeof(key->ct_labels));
305#endif /* CONFIG_NF_CONNTRACK */
306}
307EXPORT_SYMBOL(skb_flow_dissect_ct);
308
309void
310skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
311 struct flow_dissector *flow_dissector,
312 void *target_container)
313{
314 struct ip_tunnel_info *info;
315 struct ip_tunnel_key *key;
316
317 /* A quick check to see if there might be something to do. */
318 if (!dissector_uses_key(flow_dissector,
319 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
320 !dissector_uses_key(flow_dissector,
321 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
322 !dissector_uses_key(flow_dissector,
323 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
324 !dissector_uses_key(flow_dissector,
325 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
326 !dissector_uses_key(flow_dissector,
327 FLOW_DISSECTOR_KEY_ENC_PORTS) &&
328 !dissector_uses_key(flow_dissector,
329 FLOW_DISSECTOR_KEY_ENC_IP) &&
330 !dissector_uses_key(flow_dissector,
331 FLOW_DISSECTOR_KEY_ENC_OPTS))
332 return;
333
334 info = skb_tunnel_info(skb);
335 if (!info)
336 return;
337
338 key = &info->key;
339
340 switch (ip_tunnel_info_af(info)) {
341 case AF_INET:
342 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
343 flow_dissector,
344 target_container);
345 if (dissector_uses_key(flow_dissector,
346 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
347 struct flow_dissector_key_ipv4_addrs *ipv4;
348
349 ipv4 = skb_flow_dissector_target(flow_dissector,
350 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
351 target_container);
352 ipv4->src = key->u.ipv4.src;
353 ipv4->dst = key->u.ipv4.dst;
354 }
355 break;
356 case AF_INET6:
357 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
358 flow_dissector,
359 target_container);
360 if (dissector_uses_key(flow_dissector,
361 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
362 struct flow_dissector_key_ipv6_addrs *ipv6;
363
364 ipv6 = skb_flow_dissector_target(flow_dissector,
365 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
366 target_container);
367 ipv6->src = key->u.ipv6.src;
368 ipv6->dst = key->u.ipv6.dst;
369 }
370 break;
371 }
372
373 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
374 struct flow_dissector_key_keyid *keyid;
375
376 keyid = skb_flow_dissector_target(flow_dissector,
377 FLOW_DISSECTOR_KEY_ENC_KEYID,
378 target_container);
379 keyid->keyid = tunnel_id_to_key32(key->tun_id);
380 }
381
382 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
383 struct flow_dissector_key_ports *tp;
384
385 tp = skb_flow_dissector_target(flow_dissector,
386 FLOW_DISSECTOR_KEY_ENC_PORTS,
387 target_container);
388 tp->src = key->tp_src;
389 tp->dst = key->tp_dst;
390 }
391
392 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
393 struct flow_dissector_key_ip *ip;
394
395 ip = skb_flow_dissector_target(flow_dissector,
396 FLOW_DISSECTOR_KEY_ENC_IP,
397 target_container);
398 ip->tos = key->tos;
399 ip->ttl = key->ttl;
400 }
401
402 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
403 struct flow_dissector_key_enc_opts *enc_opt;
404
405 enc_opt = skb_flow_dissector_target(flow_dissector,
406 FLOW_DISSECTOR_KEY_ENC_OPTS,
407 target_container);
408
409 if (info->options_len) {
410 enc_opt->len = info->options_len;
411 ip_tunnel_info_opts_get(enc_opt->data, info);
412 enc_opt->dst_opt_type = info->key.tun_flags &
413 TUNNEL_OPTIONS_PRESENT;
414 }
415 }
416}
417EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
418
419void skb_flow_dissect_hash(const struct sk_buff *skb,
420 struct flow_dissector *flow_dissector,
421 void *target_container)
422{
423 struct flow_dissector_key_hash *key;
424
425 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_HASH))
426 return;
427
428 key = skb_flow_dissector_target(flow_dissector,
429 FLOW_DISSECTOR_KEY_HASH,
430 target_container);
431
432 key->hash = skb_get_hash_raw(skb);
433}
434EXPORT_SYMBOL(skb_flow_dissect_hash);
435
436static enum flow_dissect_ret
437__skb_flow_dissect_mpls(const struct sk_buff *skb,
438 struct flow_dissector *flow_dissector,
439 void *target_container, const void *data, int nhoff,
440 int hlen, int lse_index, bool *entropy_label)
441{
442 struct mpls_label *hdr, _hdr;
443 u32 entry, label, bos;
444
445 if (!dissector_uses_key(flow_dissector,
446 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
447 !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
448 return FLOW_DISSECT_RET_OUT_GOOD;
449
450 if (lse_index >= FLOW_DIS_MPLS_MAX)
451 return FLOW_DISSECT_RET_OUT_GOOD;
452
453 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
454 hlen, &_hdr);
455 if (!hdr)
456 return FLOW_DISSECT_RET_OUT_BAD;
457
458 entry = ntohl(hdr->entry);
459 label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
460 bos = (entry & MPLS_LS_S_MASK) >> MPLS_LS_S_SHIFT;
461
462 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
463 struct flow_dissector_key_mpls *key_mpls;
464 struct flow_dissector_mpls_lse *lse;
465
466 key_mpls = skb_flow_dissector_target(flow_dissector,
467 FLOW_DISSECTOR_KEY_MPLS,
468 target_container);
469 lse = &key_mpls->ls[lse_index];
470
471 lse->mpls_ttl = (entry & MPLS_LS_TTL_MASK) >> MPLS_LS_TTL_SHIFT;
472 lse->mpls_bos = bos;
473 lse->mpls_tc = (entry & MPLS_LS_TC_MASK) >> MPLS_LS_TC_SHIFT;
474 lse->mpls_label = label;
475 dissector_set_mpls_lse(key_mpls, lse_index);
476 }
477
478 if (*entropy_label &&
479 dissector_uses_key(flow_dissector,
480 FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
481 struct flow_dissector_key_keyid *key_keyid;
482
483 key_keyid = skb_flow_dissector_target(flow_dissector,
484 FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
485 target_container);
486 key_keyid->keyid = cpu_to_be32(label);
487 }
488
489 *entropy_label = label == MPLS_LABEL_ENTROPY;
490
491 return bos ? FLOW_DISSECT_RET_OUT_GOOD : FLOW_DISSECT_RET_PROTO_AGAIN;
492}
493
494static enum flow_dissect_ret
495__skb_flow_dissect_arp(const struct sk_buff *skb,
496 struct flow_dissector *flow_dissector,
497 void *target_container, const void *data,
498 int nhoff, int hlen)
499{
500 struct flow_dissector_key_arp *key_arp;
501 struct {
502 unsigned char ar_sha[ETH_ALEN];
503 unsigned char ar_sip[4];
504 unsigned char ar_tha[ETH_ALEN];
505 unsigned char ar_tip[4];
506 } *arp_eth, _arp_eth;
507 const struct arphdr *arp;
508 struct arphdr _arp;
509
510 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
511 return FLOW_DISSECT_RET_OUT_GOOD;
512
513 arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
514 hlen, &_arp);
515 if (!arp)
516 return FLOW_DISSECT_RET_OUT_BAD;
517
518 if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
519 arp->ar_pro != htons(ETH_P_IP) ||
520 arp->ar_hln != ETH_ALEN ||
521 arp->ar_pln != 4 ||
522 (arp->ar_op != htons(ARPOP_REPLY) &&
523 arp->ar_op != htons(ARPOP_REQUEST)))
524 return FLOW_DISSECT_RET_OUT_BAD;
525
526 arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
527 sizeof(_arp_eth), data,
528 hlen, &_arp_eth);
529 if (!arp_eth)
530 return FLOW_DISSECT_RET_OUT_BAD;
531
532 key_arp = skb_flow_dissector_target(flow_dissector,
533 FLOW_DISSECTOR_KEY_ARP,
534 target_container);
535
536 memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
537 memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
538
539 /* Only store the lower byte of the opcode;
540 * this covers ARPOP_REPLY and ARPOP_REQUEST.
541 */
542 key_arp->op = ntohs(arp->ar_op) & 0xff;
543
544 ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
545 ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
546
547 return FLOW_DISSECT_RET_OUT_GOOD;
548}
549
550static enum flow_dissect_ret
551__skb_flow_dissect_gre(const struct sk_buff *skb,
552 struct flow_dissector_key_control *key_control,
553 struct flow_dissector *flow_dissector,
554 void *target_container, const void *data,
555 __be16 *p_proto, int *p_nhoff, int *p_hlen,
556 unsigned int flags)
557{
558 struct flow_dissector_key_keyid *key_keyid;
559 struct gre_base_hdr *hdr, _hdr;
560 int offset = 0;
561 u16 gre_ver;
562
563 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
564 data, *p_hlen, &_hdr);
565 if (!hdr)
566 return FLOW_DISSECT_RET_OUT_BAD;
567
568 /* Only look inside GRE without routing */
569 if (hdr->flags & GRE_ROUTING)
570 return FLOW_DISSECT_RET_OUT_GOOD;
571
572 /* Only look inside GRE for version 0 and 1 */
573 gre_ver = ntohs(hdr->flags & GRE_VERSION);
574 if (gre_ver > 1)
575 return FLOW_DISSECT_RET_OUT_GOOD;
576
577 *p_proto = hdr->protocol;
578 if (gre_ver) {
579 /* Version1 must be PPTP, and check the flags */
580 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
581 return FLOW_DISSECT_RET_OUT_GOOD;
582 }
583
584 offset += sizeof(struct gre_base_hdr);
585
586 if (hdr->flags & GRE_CSUM)
587 offset += sizeof_field(struct gre_full_hdr, csum) +
588 sizeof_field(struct gre_full_hdr, reserved1);
589
590 if (hdr->flags & GRE_KEY) {
591 const __be32 *keyid;
592 __be32 _keyid;
593
594 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
595 sizeof(_keyid),
596 data, *p_hlen, &_keyid);
597 if (!keyid)
598 return FLOW_DISSECT_RET_OUT_BAD;
599
600 if (dissector_uses_key(flow_dissector,
601 FLOW_DISSECTOR_KEY_GRE_KEYID)) {
602 key_keyid = skb_flow_dissector_target(flow_dissector,
603 FLOW_DISSECTOR_KEY_GRE_KEYID,
604 target_container);
605 if (gre_ver == 0)
606 key_keyid->keyid = *keyid;
607 else
608 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
609 }
610 offset += sizeof_field(struct gre_full_hdr, key);
611 }
612
613 if (hdr->flags & GRE_SEQ)
614 offset += sizeof_field(struct pptp_gre_header, seq);
615
616 if (gre_ver == 0) {
617 if (*p_proto == htons(ETH_P_TEB)) {
618 const struct ethhdr *eth;
619 struct ethhdr _eth;
620
621 eth = __skb_header_pointer(skb, *p_nhoff + offset,
622 sizeof(_eth),
623 data, *p_hlen, &_eth);
624 if (!eth)
625 return FLOW_DISSECT_RET_OUT_BAD;
626 *p_proto = eth->h_proto;
627 offset += sizeof(*eth);
628
629 /* Cap headers that we access via pointers at the
630 * end of the Ethernet header as our maximum alignment
631 * at that point is only 2 bytes.
632 */
633 if (NET_IP_ALIGN)
634 *p_hlen = *p_nhoff + offset;
635 }
636 } else { /* version 1, must be PPTP */
637 u8 _ppp_hdr[PPP_HDRLEN];
638 u8 *ppp_hdr;
639
640 if (hdr->flags & GRE_ACK)
641 offset += sizeof_field(struct pptp_gre_header, ack);
642
643 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
644 sizeof(_ppp_hdr),
645 data, *p_hlen, _ppp_hdr);
646 if (!ppp_hdr)
647 return FLOW_DISSECT_RET_OUT_BAD;
648
649 switch (PPP_PROTOCOL(ppp_hdr)) {
650 case PPP_IP:
651 *p_proto = htons(ETH_P_IP);
652 break;
653 case PPP_IPV6:
654 *p_proto = htons(ETH_P_IPV6);
655 break;
656 default:
657 /* Could probably catch some more like MPLS */
658 break;
659 }
660
661 offset += PPP_HDRLEN;
662 }
663
664 *p_nhoff += offset;
665 key_control->flags |= FLOW_DIS_ENCAPSULATION;
666 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
667 return FLOW_DISSECT_RET_OUT_GOOD;
668
669 return FLOW_DISSECT_RET_PROTO_AGAIN;
670}
671
672/**
673 * __skb_flow_dissect_batadv() - dissect batman-adv header
674 * @skb: sk_buff to with the batman-adv header
675 * @key_control: flow dissectors control key
676 * @data: raw buffer pointer to the packet, if NULL use skb->data
677 * @p_proto: pointer used to update the protocol to process next
678 * @p_nhoff: pointer used to update inner network header offset
679 * @hlen: packet header length
680 * @flags: any combination of FLOW_DISSECTOR_F_*
681 *
682 * ETH_P_BATMAN packets are tried to be dissected. Only
683 * &struct batadv_unicast packets are actually processed because they contain an
684 * inner ethernet header and are usually followed by actual network header. This
685 * allows the flow dissector to continue processing the packet.
686 *
687 * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
688 * FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
689 * otherwise FLOW_DISSECT_RET_OUT_BAD
690 */
691static enum flow_dissect_ret
692__skb_flow_dissect_batadv(const struct sk_buff *skb,
693 struct flow_dissector_key_control *key_control,
694 const void *data, __be16 *p_proto, int *p_nhoff,
695 int hlen, unsigned int flags)
696{
697 struct {
698 struct batadv_unicast_packet batadv_unicast;
699 struct ethhdr eth;
700 } *hdr, _hdr;
701
702 hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
703 &_hdr);
704 if (!hdr)
705 return FLOW_DISSECT_RET_OUT_BAD;
706
707 if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
708 return FLOW_DISSECT_RET_OUT_BAD;
709
710 if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
711 return FLOW_DISSECT_RET_OUT_BAD;
712
713 *p_proto = hdr->eth.h_proto;
714 *p_nhoff += sizeof(*hdr);
715
716 key_control->flags |= FLOW_DIS_ENCAPSULATION;
717 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
718 return FLOW_DISSECT_RET_OUT_GOOD;
719
720 return FLOW_DISSECT_RET_PROTO_AGAIN;
721}
722
723static void
724__skb_flow_dissect_tcp(const struct sk_buff *skb,
725 struct flow_dissector *flow_dissector,
726 void *target_container, const void *data,
727 int thoff, int hlen)
728{
729 struct flow_dissector_key_tcp *key_tcp;
730 struct tcphdr *th, _th;
731
732 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
733 return;
734
735 th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
736 if (!th)
737 return;
738
739 if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
740 return;
741
742 key_tcp = skb_flow_dissector_target(flow_dissector,
743 FLOW_DISSECTOR_KEY_TCP,
744 target_container);
745 key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
746}
747
748static void
749__skb_flow_dissect_ports(const struct sk_buff *skb,
750 struct flow_dissector *flow_dissector,
751 void *target_container, const void *data,
752 int nhoff, u8 ip_proto, int hlen)
753{
754 enum flow_dissector_key_id dissector_ports = FLOW_DISSECTOR_KEY_MAX;
755 struct flow_dissector_key_ports *key_ports;
756
757 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
758 dissector_ports = FLOW_DISSECTOR_KEY_PORTS;
759 else if (dissector_uses_key(flow_dissector,
760 FLOW_DISSECTOR_KEY_PORTS_RANGE))
761 dissector_ports = FLOW_DISSECTOR_KEY_PORTS_RANGE;
762
763 if (dissector_ports == FLOW_DISSECTOR_KEY_MAX)
764 return;
765
766 key_ports = skb_flow_dissector_target(flow_dissector,
767 dissector_ports,
768 target_container);
769 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
770 data, hlen);
771}
772
773static void
774__skb_flow_dissect_ipv4(const struct sk_buff *skb,
775 struct flow_dissector *flow_dissector,
776 void *target_container, const void *data,
777 const struct iphdr *iph)
778{
779 struct flow_dissector_key_ip *key_ip;
780
781 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
782 return;
783
784 key_ip = skb_flow_dissector_target(flow_dissector,
785 FLOW_DISSECTOR_KEY_IP,
786 target_container);
787 key_ip->tos = iph->tos;
788 key_ip->ttl = iph->ttl;
789}
790
791static void
792__skb_flow_dissect_ipv6(const struct sk_buff *skb,
793 struct flow_dissector *flow_dissector,
794 void *target_container, const void *data,
795 const struct ipv6hdr *iph)
796{
797 struct flow_dissector_key_ip *key_ip;
798
799 if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
800 return;
801
802 key_ip = skb_flow_dissector_target(flow_dissector,
803 FLOW_DISSECTOR_KEY_IP,
804 target_container);
805 key_ip->tos = ipv6_get_dsfield(iph);
806 key_ip->ttl = iph->hop_limit;
807}
808
809/* Maximum number of protocol headers that can be parsed in
810 * __skb_flow_dissect
811 */
812#define MAX_FLOW_DISSECT_HDRS 15
813
814static bool skb_flow_dissect_allowed(int *num_hdrs)
815{
816 ++*num_hdrs;
817
818 return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
819}
820
821static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
822 struct flow_dissector *flow_dissector,
823 void *target_container)
824{
825 struct flow_dissector_key_ports *key_ports = NULL;
826 struct flow_dissector_key_control *key_control;
827 struct flow_dissector_key_basic *key_basic;
828 struct flow_dissector_key_addrs *key_addrs;
829 struct flow_dissector_key_tags *key_tags;
830
831 key_control = skb_flow_dissector_target(flow_dissector,
832 FLOW_DISSECTOR_KEY_CONTROL,
833 target_container);
834 key_control->thoff = flow_keys->thoff;
835 if (flow_keys->is_frag)
836 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
837 if (flow_keys->is_first_frag)
838 key_control->flags |= FLOW_DIS_FIRST_FRAG;
839 if (flow_keys->is_encap)
840 key_control->flags |= FLOW_DIS_ENCAPSULATION;
841
842 key_basic = skb_flow_dissector_target(flow_dissector,
843 FLOW_DISSECTOR_KEY_BASIC,
844 target_container);
845 key_basic->n_proto = flow_keys->n_proto;
846 key_basic->ip_proto = flow_keys->ip_proto;
847
848 if (flow_keys->addr_proto == ETH_P_IP &&
849 dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
850 key_addrs = skb_flow_dissector_target(flow_dissector,
851 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
852 target_container);
853 key_addrs->v4addrs.src = flow_keys->ipv4_src;
854 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
855 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
856 } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
857 dissector_uses_key(flow_dissector,
858 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
859 key_addrs = skb_flow_dissector_target(flow_dissector,
860 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
861 target_container);
862 memcpy(&key_addrs->v6addrs.src, &flow_keys->ipv6_src,
863 sizeof(key_addrs->v6addrs.src));
864 memcpy(&key_addrs->v6addrs.dst, &flow_keys->ipv6_dst,
865 sizeof(key_addrs->v6addrs.dst));
866 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
867 }
868
869 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS))
870 key_ports = skb_flow_dissector_target(flow_dissector,
871 FLOW_DISSECTOR_KEY_PORTS,
872 target_container);
873 else if (dissector_uses_key(flow_dissector,
874 FLOW_DISSECTOR_KEY_PORTS_RANGE))
875 key_ports = skb_flow_dissector_target(flow_dissector,
876 FLOW_DISSECTOR_KEY_PORTS_RANGE,
877 target_container);
878
879 if (key_ports) {
880 key_ports->src = flow_keys->sport;
881 key_ports->dst = flow_keys->dport;
882 }
883
884 if (dissector_uses_key(flow_dissector,
885 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
886 key_tags = skb_flow_dissector_target(flow_dissector,
887 FLOW_DISSECTOR_KEY_FLOW_LABEL,
888 target_container);
889 key_tags->flow_label = ntohl(flow_keys->flow_label);
890 }
891}
892
893u32 bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
894 __be16 proto, int nhoff, int hlen, unsigned int flags)
895{
896 struct bpf_flow_keys *flow_keys = ctx->flow_keys;
897 u32 result;
898
899 /* Pass parameters to the BPF program */
900 memset(flow_keys, 0, sizeof(*flow_keys));
901 flow_keys->n_proto = proto;
902 flow_keys->nhoff = nhoff;
903 flow_keys->thoff = flow_keys->nhoff;
904
905 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG !=
906 (int)FLOW_DISSECTOR_F_PARSE_1ST_FRAG);
907 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL !=
908 (int)FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
909 BUILD_BUG_ON((int)BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP !=
910 (int)FLOW_DISSECTOR_F_STOP_AT_ENCAP);
911 flow_keys->flags = flags;
912
913 result = bpf_prog_run_pin_on_cpu(prog, ctx);
914
915 flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
916 flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
917 flow_keys->nhoff, hlen);
918
919 return result;
920}
921
922static bool is_pppoe_ses_hdr_valid(const struct pppoe_hdr *hdr)
923{
924 return hdr->ver == 1 && hdr->type == 1 && hdr->code == 0;
925}
926
927/**
928 * __skb_flow_dissect - extract the flow_keys struct and return it
929 * @net: associated network namespace, derived from @skb if NULL
930 * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
931 * @flow_dissector: list of keys to dissect
932 * @target_container: target structure to put dissected values into
933 * @data: raw buffer pointer to the packet, if NULL use skb->data
934 * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
935 * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
936 * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
937 * @flags: flags that control the dissection process, e.g.
938 * FLOW_DISSECTOR_F_STOP_AT_ENCAP.
939 *
940 * The function will try to retrieve individual keys into target specified
941 * by flow_dissector from either the skbuff or a raw buffer specified by the
942 * rest parameters.
943 *
944 * Caller must take care of zeroing target container memory.
945 */
946bool __skb_flow_dissect(const struct net *net,
947 const struct sk_buff *skb,
948 struct flow_dissector *flow_dissector,
949 void *target_container, const void *data,
950 __be16 proto, int nhoff, int hlen, unsigned int flags)
951{
952 struct flow_dissector_key_control *key_control;
953 struct flow_dissector_key_basic *key_basic;
954 struct flow_dissector_key_addrs *key_addrs;
955 struct flow_dissector_key_tags *key_tags;
956 struct flow_dissector_key_vlan *key_vlan;
957 enum flow_dissect_ret fdret;
958 enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
959 bool mpls_el = false;
960 int mpls_lse = 0;
961 int num_hdrs = 0;
962 u8 ip_proto = 0;
963 bool ret;
964
965 if (!data) {
966 data = skb->data;
967 proto = skb_vlan_tag_present(skb) ?
968 skb->vlan_proto : skb->protocol;
969 nhoff = skb_network_offset(skb);
970 hlen = skb_headlen(skb);
971#if IS_ENABLED(CONFIG_NET_DSA)
972 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev) &&
973 proto == htons(ETH_P_XDSA))) {
974 struct metadata_dst *md_dst = skb_metadata_dst(skb);
975 const struct dsa_device_ops *ops;
976 int offset = 0;
977
978 ops = skb->dev->dsa_ptr->tag_ops;
979 /* Only DSA header taggers break flow dissection */
980 if (ops->needed_headroom &&
981 (!md_dst || md_dst->type != METADATA_HW_PORT_MUX)) {
982 if (ops->flow_dissect)
983 ops->flow_dissect(skb, &proto, &offset);
984 else
985 dsa_tag_generic_flow_dissect(skb,
986 &proto,
987 &offset);
988 hlen -= offset;
989 nhoff += offset;
990 }
991 }
992#endif
993 }
994
995 /* It is ensured by skb_flow_dissector_init() that control key will
996 * be always present.
997 */
998 key_control = skb_flow_dissector_target(flow_dissector,
999 FLOW_DISSECTOR_KEY_CONTROL,
1000 target_container);
1001
1002 /* It is ensured by skb_flow_dissector_init() that basic key will
1003 * be always present.
1004 */
1005 key_basic = skb_flow_dissector_target(flow_dissector,
1006 FLOW_DISSECTOR_KEY_BASIC,
1007 target_container);
1008
1009 if (skb) {
1010 if (!net) {
1011 if (skb->dev)
1012 net = dev_net(skb->dev);
1013 else if (skb->sk)
1014 net = sock_net(skb->sk);
1015 }
1016 }
1017
1018 WARN_ON_ONCE(!net);
1019 if (net) {
1020 enum netns_bpf_attach_type type = NETNS_BPF_FLOW_DISSECTOR;
1021 struct bpf_prog_array *run_array;
1022
1023 rcu_read_lock();
1024 run_array = rcu_dereference(init_net.bpf.run_array[type]);
1025 if (!run_array)
1026 run_array = rcu_dereference(net->bpf.run_array[type]);
1027
1028 if (run_array) {
1029 struct bpf_flow_keys flow_keys;
1030 struct bpf_flow_dissector ctx = {
1031 .flow_keys = &flow_keys,
1032 .data = data,
1033 .data_end = data + hlen,
1034 };
1035 __be16 n_proto = proto;
1036 struct bpf_prog *prog;
1037 u32 result;
1038
1039 if (skb) {
1040 ctx.skb = skb;
1041 /* we can't use 'proto' in the skb case
1042 * because it might be set to skb->vlan_proto
1043 * which has been pulled from the data
1044 */
1045 n_proto = skb->protocol;
1046 }
1047
1048 prog = READ_ONCE(run_array->items[0].prog);
1049 result = bpf_flow_dissect(prog, &ctx, n_proto, nhoff,
1050 hlen, flags);
1051 if (result == BPF_FLOW_DISSECTOR_CONTINUE)
1052 goto dissect_continue;
1053 __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
1054 target_container);
1055 rcu_read_unlock();
1056 return result == BPF_OK;
1057 }
1058dissect_continue:
1059 rcu_read_unlock();
1060 }
1061
1062 if (dissector_uses_key(flow_dissector,
1063 FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
1064 struct ethhdr *eth = eth_hdr(skb);
1065 struct flow_dissector_key_eth_addrs *key_eth_addrs;
1066
1067 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
1068 FLOW_DISSECTOR_KEY_ETH_ADDRS,
1069 target_container);
1070 memcpy(key_eth_addrs, eth, sizeof(*key_eth_addrs));
1071 }
1072
1073 if (dissector_uses_key(flow_dissector,
1074 FLOW_DISSECTOR_KEY_NUM_OF_VLANS)) {
1075 struct flow_dissector_key_num_of_vlans *key_num_of_vlans;
1076
1077 key_num_of_vlans = skb_flow_dissector_target(flow_dissector,
1078 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1079 target_container);
1080 key_num_of_vlans->num_of_vlans = 0;
1081 }
1082
1083proto_again:
1084 fdret = FLOW_DISSECT_RET_CONTINUE;
1085
1086 switch (proto) {
1087 case htons(ETH_P_IP): {
1088 const struct iphdr *iph;
1089 struct iphdr _iph;
1090
1091 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1092 if (!iph || iph->ihl < 5) {
1093 fdret = FLOW_DISSECT_RET_OUT_BAD;
1094 break;
1095 }
1096
1097 nhoff += iph->ihl * 4;
1098
1099 ip_proto = iph->protocol;
1100
1101 if (dissector_uses_key(flow_dissector,
1102 FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
1103 key_addrs = skb_flow_dissector_target(flow_dissector,
1104 FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1105 target_container);
1106
1107 memcpy(&key_addrs->v4addrs.src, &iph->saddr,
1108 sizeof(key_addrs->v4addrs.src));
1109 memcpy(&key_addrs->v4addrs.dst, &iph->daddr,
1110 sizeof(key_addrs->v4addrs.dst));
1111 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
1112 }
1113
1114 __skb_flow_dissect_ipv4(skb, flow_dissector,
1115 target_container, data, iph);
1116
1117 if (ip_is_fragment(iph)) {
1118 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1119
1120 if (iph->frag_off & htons(IP_OFFSET)) {
1121 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1122 break;
1123 } else {
1124 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1125 if (!(flags &
1126 FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
1127 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1128 break;
1129 }
1130 }
1131 }
1132
1133 break;
1134 }
1135 case htons(ETH_P_IPV6): {
1136 const struct ipv6hdr *iph;
1137 struct ipv6hdr _iph;
1138
1139 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
1140 if (!iph) {
1141 fdret = FLOW_DISSECT_RET_OUT_BAD;
1142 break;
1143 }
1144
1145 ip_proto = iph->nexthdr;
1146 nhoff += sizeof(struct ipv6hdr);
1147
1148 if (dissector_uses_key(flow_dissector,
1149 FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
1150 key_addrs = skb_flow_dissector_target(flow_dissector,
1151 FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1152 target_container);
1153
1154 memcpy(&key_addrs->v6addrs.src, &iph->saddr,
1155 sizeof(key_addrs->v6addrs.src));
1156 memcpy(&key_addrs->v6addrs.dst, &iph->daddr,
1157 sizeof(key_addrs->v6addrs.dst));
1158 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1159 }
1160
1161 if ((dissector_uses_key(flow_dissector,
1162 FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
1163 (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
1164 ip6_flowlabel(iph)) {
1165 __be32 flow_label = ip6_flowlabel(iph);
1166
1167 if (dissector_uses_key(flow_dissector,
1168 FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
1169 key_tags = skb_flow_dissector_target(flow_dissector,
1170 FLOW_DISSECTOR_KEY_FLOW_LABEL,
1171 target_container);
1172 key_tags->flow_label = ntohl(flow_label);
1173 }
1174 if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
1175 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1176 break;
1177 }
1178 }
1179
1180 __skb_flow_dissect_ipv6(skb, flow_dissector,
1181 target_container, data, iph);
1182
1183 break;
1184 }
1185 case htons(ETH_P_8021AD):
1186 case htons(ETH_P_8021Q): {
1187 const struct vlan_hdr *vlan = NULL;
1188 struct vlan_hdr _vlan;
1189 __be16 saved_vlan_tpid = proto;
1190
1191 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
1192 skb && skb_vlan_tag_present(skb)) {
1193 proto = skb->protocol;
1194 } else {
1195 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
1196 data, hlen, &_vlan);
1197 if (!vlan) {
1198 fdret = FLOW_DISSECT_RET_OUT_BAD;
1199 break;
1200 }
1201
1202 proto = vlan->h_vlan_encapsulated_proto;
1203 nhoff += sizeof(*vlan);
1204 }
1205
1206 if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_NUM_OF_VLANS) &&
1207 !(key_control->flags & FLOW_DIS_ENCAPSULATION)) {
1208 struct flow_dissector_key_num_of_vlans *key_nvs;
1209
1210 key_nvs = skb_flow_dissector_target(flow_dissector,
1211 FLOW_DISSECTOR_KEY_NUM_OF_VLANS,
1212 target_container);
1213 key_nvs->num_of_vlans++;
1214 }
1215
1216 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1217 dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1218 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1219 dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1220 } else {
1221 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1222 break;
1223 }
1224
1225 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1226 key_vlan = skb_flow_dissector_target(flow_dissector,
1227 dissector_vlan,
1228 target_container);
1229
1230 if (!vlan) {
1231 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1232 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1233 } else {
1234 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1235 VLAN_VID_MASK;
1236 key_vlan->vlan_priority =
1237 (ntohs(vlan->h_vlan_TCI) &
1238 VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1239 }
1240 key_vlan->vlan_tpid = saved_vlan_tpid;
1241 key_vlan->vlan_eth_type = proto;
1242 }
1243
1244 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1245 break;
1246 }
1247 case htons(ETH_P_PPP_SES): {
1248 struct {
1249 struct pppoe_hdr hdr;
1250 __be16 proto;
1251 } *hdr, _hdr;
1252 u16 ppp_proto;
1253
1254 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1255 if (!hdr) {
1256 fdret = FLOW_DISSECT_RET_OUT_BAD;
1257 break;
1258 }
1259
1260 if (!is_pppoe_ses_hdr_valid(&hdr->hdr)) {
1261 fdret = FLOW_DISSECT_RET_OUT_BAD;
1262 break;
1263 }
1264
1265 /* least significant bit of the most significant octet
1266 * indicates if protocol field was compressed
1267 */
1268 ppp_proto = ntohs(hdr->proto);
1269 if (ppp_proto & 0x0100) {
1270 ppp_proto = ppp_proto >> 8;
1271 nhoff += PPPOE_SES_HLEN - 1;
1272 } else {
1273 nhoff += PPPOE_SES_HLEN;
1274 }
1275
1276 if (ppp_proto == PPP_IP) {
1277 proto = htons(ETH_P_IP);
1278 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1279 } else if (ppp_proto == PPP_IPV6) {
1280 proto = htons(ETH_P_IPV6);
1281 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1282 } else if (ppp_proto == PPP_MPLS_UC) {
1283 proto = htons(ETH_P_MPLS_UC);
1284 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1285 } else if (ppp_proto == PPP_MPLS_MC) {
1286 proto = htons(ETH_P_MPLS_MC);
1287 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1288 } else if (ppp_proto_is_valid(ppp_proto)) {
1289 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1290 } else {
1291 fdret = FLOW_DISSECT_RET_OUT_BAD;
1292 break;
1293 }
1294
1295 if (dissector_uses_key(flow_dissector,
1296 FLOW_DISSECTOR_KEY_PPPOE)) {
1297 struct flow_dissector_key_pppoe *key_pppoe;
1298
1299 key_pppoe = skb_flow_dissector_target(flow_dissector,
1300 FLOW_DISSECTOR_KEY_PPPOE,
1301 target_container);
1302 key_pppoe->session_id = hdr->hdr.sid;
1303 key_pppoe->ppp_proto = htons(ppp_proto);
1304 key_pppoe->type = htons(ETH_P_PPP_SES);
1305 }
1306 break;
1307 }
1308 case htons(ETH_P_TIPC): {
1309 struct tipc_basic_hdr *hdr, _hdr;
1310
1311 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1312 data, hlen, &_hdr);
1313 if (!hdr) {
1314 fdret = FLOW_DISSECT_RET_OUT_BAD;
1315 break;
1316 }
1317
1318 if (dissector_uses_key(flow_dissector,
1319 FLOW_DISSECTOR_KEY_TIPC)) {
1320 key_addrs = skb_flow_dissector_target(flow_dissector,
1321 FLOW_DISSECTOR_KEY_TIPC,
1322 target_container);
1323 key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1324 key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1325 }
1326 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1327 break;
1328 }
1329
1330 case htons(ETH_P_MPLS_UC):
1331 case htons(ETH_P_MPLS_MC):
1332 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1333 target_container, data,
1334 nhoff, hlen, mpls_lse,
1335 &mpls_el);
1336 nhoff += sizeof(struct mpls_label);
1337 mpls_lse++;
1338 break;
1339 case htons(ETH_P_FCOE):
1340 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1341 fdret = FLOW_DISSECT_RET_OUT_BAD;
1342 break;
1343 }
1344
1345 nhoff += FCOE_HEADER_LEN;
1346 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1347 break;
1348
1349 case htons(ETH_P_ARP):
1350 case htons(ETH_P_RARP):
1351 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1352 target_container, data,
1353 nhoff, hlen);
1354 break;
1355
1356 case htons(ETH_P_BATMAN):
1357 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1358 &proto, &nhoff, hlen, flags);
1359 break;
1360
1361 case htons(ETH_P_1588): {
1362 struct ptp_header *hdr, _hdr;
1363
1364 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
1365 hlen, &_hdr);
1366 if (!hdr) {
1367 fdret = FLOW_DISSECT_RET_OUT_BAD;
1368 break;
1369 }
1370
1371 nhoff += ntohs(hdr->message_length);
1372 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1373 break;
1374 }
1375
1376 case htons(ETH_P_PRP):
1377 case htons(ETH_P_HSR): {
1378 struct hsr_tag *hdr, _hdr;
1379
1380 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen,
1381 &_hdr);
1382 if (!hdr) {
1383 fdret = FLOW_DISSECT_RET_OUT_BAD;
1384 break;
1385 }
1386
1387 proto = hdr->encap_proto;
1388 nhoff += HSR_HLEN;
1389 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1390 break;
1391 }
1392
1393 default:
1394 fdret = FLOW_DISSECT_RET_OUT_BAD;
1395 break;
1396 }
1397
1398 /* Process result of proto processing */
1399 switch (fdret) {
1400 case FLOW_DISSECT_RET_OUT_GOOD:
1401 goto out_good;
1402 case FLOW_DISSECT_RET_PROTO_AGAIN:
1403 if (skb_flow_dissect_allowed(&num_hdrs))
1404 goto proto_again;
1405 goto out_good;
1406 case FLOW_DISSECT_RET_CONTINUE:
1407 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1408 break;
1409 case FLOW_DISSECT_RET_OUT_BAD:
1410 default:
1411 goto out_bad;
1412 }
1413
1414ip_proto_again:
1415 fdret = FLOW_DISSECT_RET_CONTINUE;
1416
1417 switch (ip_proto) {
1418 case IPPROTO_GRE:
1419 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1420 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1421 break;
1422 }
1423
1424 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1425 target_container, data,
1426 &proto, &nhoff, &hlen, flags);
1427 break;
1428
1429 case NEXTHDR_HOP:
1430 case NEXTHDR_ROUTING:
1431 case NEXTHDR_DEST: {
1432 u8 _opthdr[2], *opthdr;
1433
1434 if (proto != htons(ETH_P_IPV6))
1435 break;
1436
1437 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1438 data, hlen, &_opthdr);
1439 if (!opthdr) {
1440 fdret = FLOW_DISSECT_RET_OUT_BAD;
1441 break;
1442 }
1443
1444 ip_proto = opthdr[0];
1445 nhoff += (opthdr[1] + 1) << 3;
1446
1447 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1448 break;
1449 }
1450 case NEXTHDR_FRAGMENT: {
1451 struct frag_hdr _fh, *fh;
1452
1453 if (proto != htons(ETH_P_IPV6))
1454 break;
1455
1456 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1457 data, hlen, &_fh);
1458
1459 if (!fh) {
1460 fdret = FLOW_DISSECT_RET_OUT_BAD;
1461 break;
1462 }
1463
1464 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1465
1466 nhoff += sizeof(_fh);
1467 ip_proto = fh->nexthdr;
1468
1469 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1470 key_control->flags |= FLOW_DIS_FIRST_FRAG;
1471 if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1472 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1473 break;
1474 }
1475 }
1476
1477 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1478 break;
1479 }
1480 case IPPROTO_IPIP:
1481 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1482 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1483 break;
1484 }
1485
1486 proto = htons(ETH_P_IP);
1487
1488 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1489 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1490 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1491 break;
1492 }
1493
1494 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1495 break;
1496
1497 case IPPROTO_IPV6:
1498 if (flags & FLOW_DISSECTOR_F_STOP_BEFORE_ENCAP) {
1499 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1500 break;
1501 }
1502
1503 proto = htons(ETH_P_IPV6);
1504
1505 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1506 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1507 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1508 break;
1509 }
1510
1511 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1512 break;
1513
1514
1515 case IPPROTO_MPLS:
1516 proto = htons(ETH_P_MPLS_UC);
1517 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1518 break;
1519
1520 case IPPROTO_TCP:
1521 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1522 data, nhoff, hlen);
1523 break;
1524
1525 case IPPROTO_ICMP:
1526 case IPPROTO_ICMPV6:
1527 __skb_flow_dissect_icmp(skb, flow_dissector, target_container,
1528 data, nhoff, hlen);
1529 break;
1530 case IPPROTO_L2TP:
1531 __skb_flow_dissect_l2tpv3(skb, flow_dissector, target_container,
1532 data, nhoff, hlen);
1533 break;
1534
1535 default:
1536 break;
1537 }
1538
1539 if (!(key_control->flags & FLOW_DIS_IS_FRAGMENT))
1540 __skb_flow_dissect_ports(skb, flow_dissector, target_container,
1541 data, nhoff, ip_proto, hlen);
1542
1543 /* Process result of IP proto processing */
1544 switch (fdret) {
1545 case FLOW_DISSECT_RET_PROTO_AGAIN:
1546 if (skb_flow_dissect_allowed(&num_hdrs))
1547 goto proto_again;
1548 break;
1549 case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1550 if (skb_flow_dissect_allowed(&num_hdrs))
1551 goto ip_proto_again;
1552 break;
1553 case FLOW_DISSECT_RET_OUT_GOOD:
1554 case FLOW_DISSECT_RET_CONTINUE:
1555 break;
1556 case FLOW_DISSECT_RET_OUT_BAD:
1557 default:
1558 goto out_bad;
1559 }
1560
1561out_good:
1562 ret = true;
1563
1564out:
1565 key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1566 key_basic->n_proto = proto;
1567 key_basic->ip_proto = ip_proto;
1568
1569 return ret;
1570
1571out_bad:
1572 ret = false;
1573 goto out;
1574}
1575EXPORT_SYMBOL(__skb_flow_dissect);
1576
1577static siphash_aligned_key_t hashrnd;
1578static __always_inline void __flow_hash_secret_init(void)
1579{
1580 net_get_random_once(&hashrnd, sizeof(hashrnd));
1581}
1582
1583static const void *flow_keys_hash_start(const struct flow_keys *flow)
1584{
1585 BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % SIPHASH_ALIGNMENT);
1586 return &flow->FLOW_KEYS_HASH_START_FIELD;
1587}
1588
1589static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1590{
1591 size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1592
1593 BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1594
1595 switch (flow->control.addr_type) {
1596 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1597 diff -= sizeof(flow->addrs.v4addrs);
1598 break;
1599 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1600 diff -= sizeof(flow->addrs.v6addrs);
1601 break;
1602 case FLOW_DISSECTOR_KEY_TIPC:
1603 diff -= sizeof(flow->addrs.tipckey);
1604 break;
1605 }
1606 return sizeof(*flow) - diff;
1607}
1608
1609__be32 flow_get_u32_src(const struct flow_keys *flow)
1610{
1611 switch (flow->control.addr_type) {
1612 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1613 return flow->addrs.v4addrs.src;
1614 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1615 return (__force __be32)ipv6_addr_hash(
1616 &flow->addrs.v6addrs.src);
1617 case FLOW_DISSECTOR_KEY_TIPC:
1618 return flow->addrs.tipckey.key;
1619 default:
1620 return 0;
1621 }
1622}
1623EXPORT_SYMBOL(flow_get_u32_src);
1624
1625__be32 flow_get_u32_dst(const struct flow_keys *flow)
1626{
1627 switch (flow->control.addr_type) {
1628 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1629 return flow->addrs.v4addrs.dst;
1630 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1631 return (__force __be32)ipv6_addr_hash(
1632 &flow->addrs.v6addrs.dst);
1633 default:
1634 return 0;
1635 }
1636}
1637EXPORT_SYMBOL(flow_get_u32_dst);
1638
1639/* Sort the source and destination IP and the ports,
1640 * to have consistent hash within the two directions
1641 */
1642static inline void __flow_hash_consistentify(struct flow_keys *keys)
1643{
1644 int addr_diff, i;
1645
1646 switch (keys->control.addr_type) {
1647 case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1648 if ((__force u32)keys->addrs.v4addrs.dst <
1649 (__force u32)keys->addrs.v4addrs.src)
1650 swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1651
1652 if ((__force u16)keys->ports.dst <
1653 (__force u16)keys->ports.src) {
1654 swap(keys->ports.src, keys->ports.dst);
1655 }
1656 break;
1657 case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1658 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1659 &keys->addrs.v6addrs.src,
1660 sizeof(keys->addrs.v6addrs.dst));
1661 if (addr_diff < 0) {
1662 for (i = 0; i < 4; i++)
1663 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1664 keys->addrs.v6addrs.dst.s6_addr32[i]);
1665 }
1666 if ((__force u16)keys->ports.dst <
1667 (__force u16)keys->ports.src) {
1668 swap(keys->ports.src, keys->ports.dst);
1669 }
1670 break;
1671 }
1672}
1673
1674static inline u32 __flow_hash_from_keys(struct flow_keys *keys,
1675 const siphash_key_t *keyval)
1676{
1677 u32 hash;
1678
1679 __flow_hash_consistentify(keys);
1680
1681 hash = siphash(flow_keys_hash_start(keys),
1682 flow_keys_hash_length(keys), keyval);
1683 if (!hash)
1684 hash = 1;
1685
1686 return hash;
1687}
1688
1689u32 flow_hash_from_keys(struct flow_keys *keys)
1690{
1691 __flow_hash_secret_init();
1692 return __flow_hash_from_keys(keys, &hashrnd);
1693}
1694EXPORT_SYMBOL(flow_hash_from_keys);
1695
1696static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1697 struct flow_keys *keys,
1698 const siphash_key_t *keyval)
1699{
1700 skb_flow_dissect_flow_keys(skb, keys,
1701 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1702
1703 return __flow_hash_from_keys(keys, keyval);
1704}
1705
1706struct _flow_keys_digest_data {
1707 __be16 n_proto;
1708 u8 ip_proto;
1709 u8 padding;
1710 __be32 ports;
1711 __be32 src;
1712 __be32 dst;
1713};
1714
1715void make_flow_keys_digest(struct flow_keys_digest *digest,
1716 const struct flow_keys *flow)
1717{
1718 struct _flow_keys_digest_data *data =
1719 (struct _flow_keys_digest_data *)digest;
1720
1721 BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1722
1723 memset(digest, 0, sizeof(*digest));
1724
1725 data->n_proto = flow->basic.n_proto;
1726 data->ip_proto = flow->basic.ip_proto;
1727 data->ports = flow->ports.ports;
1728 data->src = flow->addrs.v4addrs.src;
1729 data->dst = flow->addrs.v4addrs.dst;
1730}
1731EXPORT_SYMBOL(make_flow_keys_digest);
1732
1733static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1734
1735u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1736{
1737 struct flow_keys keys;
1738
1739 __flow_hash_secret_init();
1740
1741 memset(&keys, 0, sizeof(keys));
1742 __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1743 &keys, NULL, 0, 0, 0,
1744 FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1745
1746 return __flow_hash_from_keys(&keys, &hashrnd);
1747}
1748EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1749
1750/**
1751 * __skb_get_hash: calculate a flow hash
1752 * @skb: sk_buff to calculate flow hash from
1753 *
1754 * This function calculates a flow hash based on src/dst addresses
1755 * and src/dst port numbers. Sets hash in skb to non-zero hash value
1756 * on success, zero indicates no valid hash. Also, sets l4_hash in skb
1757 * if hash is a canonical 4-tuple hash over transport ports.
1758 */
1759void __skb_get_hash(struct sk_buff *skb)
1760{
1761 struct flow_keys keys;
1762 u32 hash;
1763
1764 __flow_hash_secret_init();
1765
1766 hash = ___skb_get_hash(skb, &keys, &hashrnd);
1767
1768 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1769}
1770EXPORT_SYMBOL(__skb_get_hash);
1771
1772__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1773 const siphash_key_t *perturb)
1774{
1775 struct flow_keys keys;
1776
1777 return ___skb_get_hash(skb, &keys, perturb);
1778}
1779EXPORT_SYMBOL(skb_get_hash_perturb);
1780
1781u32 __skb_get_poff(const struct sk_buff *skb, const void *data,
1782 const struct flow_keys_basic *keys, int hlen)
1783{
1784 u32 poff = keys->control.thoff;
1785
1786 /* skip L4 headers for fragments after the first */
1787 if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1788 !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1789 return poff;
1790
1791 switch (keys->basic.ip_proto) {
1792 case IPPROTO_TCP: {
1793 /* access doff as u8 to avoid unaligned access */
1794 const u8 *doff;
1795 u8 _doff;
1796
1797 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1798 data, hlen, &_doff);
1799 if (!doff)
1800 return poff;
1801
1802 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1803 break;
1804 }
1805 case IPPROTO_UDP:
1806 case IPPROTO_UDPLITE:
1807 poff += sizeof(struct udphdr);
1808 break;
1809 /* For the rest, we do not really care about header
1810 * extensions at this point for now.
1811 */
1812 case IPPROTO_ICMP:
1813 poff += sizeof(struct icmphdr);
1814 break;
1815 case IPPROTO_ICMPV6:
1816 poff += sizeof(struct icmp6hdr);
1817 break;
1818 case IPPROTO_IGMP:
1819 poff += sizeof(struct igmphdr);
1820 break;
1821 case IPPROTO_DCCP:
1822 poff += sizeof(struct dccp_hdr);
1823 break;
1824 case IPPROTO_SCTP:
1825 poff += sizeof(struct sctphdr);
1826 break;
1827 }
1828
1829 return poff;
1830}
1831
1832/**
1833 * skb_get_poff - get the offset to the payload
1834 * @skb: sk_buff to get the payload offset from
1835 *
1836 * The function will get the offset to the payload as far as it could
1837 * be dissected. The main user is currently BPF, so that we can dynamically
1838 * truncate packets without needing to push actual payload to the user
1839 * space and can analyze headers only, instead.
1840 */
1841u32 skb_get_poff(const struct sk_buff *skb)
1842{
1843 struct flow_keys_basic keys;
1844
1845 if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1846 NULL, 0, 0, 0, 0))
1847 return 0;
1848
1849 return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1850}
1851
1852__u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1853{
1854 memset(keys, 0, sizeof(*keys));
1855
1856 memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1857 sizeof(keys->addrs.v6addrs.src));
1858 memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1859 sizeof(keys->addrs.v6addrs.dst));
1860 keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1861 keys->ports.src = fl6->fl6_sport;
1862 keys->ports.dst = fl6->fl6_dport;
1863 keys->keyid.keyid = fl6->fl6_gre_key;
1864 keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1865 keys->basic.ip_proto = fl6->flowi6_proto;
1866
1867 return flow_hash_from_keys(keys);
1868}
1869EXPORT_SYMBOL(__get_hash_from_flowi6);
1870
1871static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1872 {
1873 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1874 .offset = offsetof(struct flow_keys, control),
1875 },
1876 {
1877 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1878 .offset = offsetof(struct flow_keys, basic),
1879 },
1880 {
1881 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1882 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1883 },
1884 {
1885 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1886 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1887 },
1888 {
1889 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1890 .offset = offsetof(struct flow_keys, addrs.tipckey),
1891 },
1892 {
1893 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1894 .offset = offsetof(struct flow_keys, ports),
1895 },
1896 {
1897 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1898 .offset = offsetof(struct flow_keys, vlan),
1899 },
1900 {
1901 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1902 .offset = offsetof(struct flow_keys, tags),
1903 },
1904 {
1905 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1906 .offset = offsetof(struct flow_keys, keyid),
1907 },
1908};
1909
1910static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1911 {
1912 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1913 .offset = offsetof(struct flow_keys, control),
1914 },
1915 {
1916 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1917 .offset = offsetof(struct flow_keys, basic),
1918 },
1919 {
1920 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1921 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1922 },
1923 {
1924 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1925 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1926 },
1927 {
1928 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1929 .offset = offsetof(struct flow_keys, ports),
1930 },
1931};
1932
1933static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1934 {
1935 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1936 .offset = offsetof(struct flow_keys, control),
1937 },
1938 {
1939 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1940 .offset = offsetof(struct flow_keys, basic),
1941 },
1942};
1943
1944struct flow_dissector flow_keys_dissector __read_mostly;
1945EXPORT_SYMBOL(flow_keys_dissector);
1946
1947struct flow_dissector flow_keys_basic_dissector __read_mostly;
1948EXPORT_SYMBOL(flow_keys_basic_dissector);
1949
1950static int __init init_default_flow_dissectors(void)
1951{
1952 skb_flow_dissector_init(&flow_keys_dissector,
1953 flow_keys_dissector_keys,
1954 ARRAY_SIZE(flow_keys_dissector_keys));
1955 skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1956 flow_keys_dissector_symmetric_keys,
1957 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1958 skb_flow_dissector_init(&flow_keys_basic_dissector,
1959 flow_keys_basic_dissector_keys,
1960 ARRAY_SIZE(flow_keys_basic_dissector_keys));
1961 return 0;
1962}
1963core_initcall(init_default_flow_dissectors);