flow_dissector: Correctly handle parsing FCoE
[cascardo/linux.git] / net / core / flow_dissector.c
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/ip.h>
8 #include <net/ipv6.h>
9 #include <linux/igmp.h>
10 #include <linux/icmp.h>
11 #include <linux/sctp.h>
12 #include <linux/dccp.h>
13 #include <linux/if_tunnel.h>
14 #include <linux/if_pppox.h>
15 #include <linux/ppp_defs.h>
16 #include <linux/stddef.h>
17 #include <linux/if_ether.h>
18 #include <linux/mpls.h>
19 #include <net/flow_dissector.h>
20 #include <scsi/fc/fc_fcoe.h>
21
22 static bool dissector_uses_key(const struct flow_dissector *flow_dissector,
23                                enum flow_dissector_key_id key_id)
24 {
25         return flow_dissector->used_keys & (1 << key_id);
26 }
27
28 static void dissector_set_key(struct flow_dissector *flow_dissector,
29                               enum flow_dissector_key_id key_id)
30 {
31         flow_dissector->used_keys |= (1 << key_id);
32 }
33
34 static void *skb_flow_dissector_target(struct flow_dissector *flow_dissector,
35                                        enum flow_dissector_key_id key_id,
36                                        void *target_container)
37 {
38         return ((char *) target_container) + flow_dissector->offset[key_id];
39 }
40
41 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
42                              const struct flow_dissector_key *key,
43                              unsigned int key_count)
44 {
45         unsigned int i;
46
47         memset(flow_dissector, 0, sizeof(*flow_dissector));
48
49         for (i = 0; i < key_count; i++, key++) {
50                 /* User should make sure that every key target offset is withing
51                  * boundaries of unsigned short.
52                  */
53                 BUG_ON(key->offset > USHRT_MAX);
54                 BUG_ON(dissector_uses_key(flow_dissector,
55                                           key->key_id));
56
57                 dissector_set_key(flow_dissector, key->key_id);
58                 flow_dissector->offset[key->key_id] = key->offset;
59         }
60
61         /* Ensure that the dissector always includes control and basic key.
62          * That way we are able to avoid handling lack of these in fast path.
63          */
64         BUG_ON(!dissector_uses_key(flow_dissector,
65                                    FLOW_DISSECTOR_KEY_CONTROL));
66         BUG_ON(!dissector_uses_key(flow_dissector,
67                                    FLOW_DISSECTOR_KEY_BASIC));
68 }
69 EXPORT_SYMBOL(skb_flow_dissector_init);
70
71 /**
72  * __skb_flow_get_ports - extract the upper layer ports and return them
73  * @skb: sk_buff to extract the ports from
74  * @thoff: transport header offset
75  * @ip_proto: protocol for which to get port offset
76  * @data: raw buffer pointer to the packet, if NULL use skb->data
77  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
78  *
79  * The function will try to retrieve the ports at offset thoff + poff where poff
80  * is the protocol port offset returned from proto_ports_offset
81  */
82 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
83                             void *data, int hlen)
84 {
85         int poff = proto_ports_offset(ip_proto);
86
87         if (!data) {
88                 data = skb->data;
89                 hlen = skb_headlen(skb);
90         }
91
92         if (poff >= 0) {
93                 __be32 *ports, _ports;
94
95                 ports = __skb_header_pointer(skb, thoff + poff,
96                                              sizeof(_ports), data, hlen, &_ports);
97                 if (ports)
98                         return *ports;
99         }
100
101         return 0;
102 }
103 EXPORT_SYMBOL(__skb_flow_get_ports);
104
105 /**
106  * __skb_flow_dissect - extract the flow_keys struct and return it
107  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
108  * @flow_dissector: list of keys to dissect
109  * @target_container: target structure to put dissected values into
110  * @data: raw buffer pointer to the packet, if NULL use skb->data
111  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
112  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
113  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
114  *
115  * The function will try to retrieve individual keys into target specified
116  * by flow_dissector from either the skbuff or a raw buffer specified by the
117  * rest parameters.
118  *
119  * Caller must take care of zeroing target container memory.
120  */
121 bool __skb_flow_dissect(const struct sk_buff *skb,
122                         struct flow_dissector *flow_dissector,
123                         void *target_container,
124                         void *data, __be16 proto, int nhoff, int hlen,
125                         unsigned int flags)
126 {
127         struct flow_dissector_key_control *key_control;
128         struct flow_dissector_key_basic *key_basic;
129         struct flow_dissector_key_addrs *key_addrs;
130         struct flow_dissector_key_ports *key_ports;
131         struct flow_dissector_key_tags *key_tags;
132         struct flow_dissector_key_keyid *key_keyid;
133         u8 ip_proto = 0;
134         bool ret = false;
135
136         if (!data) {
137                 data = skb->data;
138                 proto = skb->protocol;
139                 nhoff = skb_network_offset(skb);
140                 hlen = skb_headlen(skb);
141         }
142
143         /* It is ensured by skb_flow_dissector_init() that control key will
144          * be always present.
145          */
146         key_control = skb_flow_dissector_target(flow_dissector,
147                                                 FLOW_DISSECTOR_KEY_CONTROL,
148                                                 target_container);
149
150         /* It is ensured by skb_flow_dissector_init() that basic key will
151          * be always present.
152          */
153         key_basic = skb_flow_dissector_target(flow_dissector,
154                                               FLOW_DISSECTOR_KEY_BASIC,
155                                               target_container);
156
157         if (dissector_uses_key(flow_dissector,
158                                FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
159                 struct ethhdr *eth = eth_hdr(skb);
160                 struct flow_dissector_key_eth_addrs *key_eth_addrs;
161
162                 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
163                                                           FLOW_DISSECTOR_KEY_ETH_ADDRS,
164                                                           target_container);
165                 memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
166         }
167
168 again:
169         switch (proto) {
170         case htons(ETH_P_IP): {
171                 const struct iphdr *iph;
172                 struct iphdr _iph;
173 ip:
174                 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
175                 if (!iph || iph->ihl < 5)
176                         goto out_bad;
177                 nhoff += iph->ihl * 4;
178
179                 ip_proto = iph->protocol;
180
181                 if (dissector_uses_key(flow_dissector,
182                                        FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
183                         key_addrs = skb_flow_dissector_target(flow_dissector,
184                                                               FLOW_DISSECTOR_KEY_IPV4_ADDRS,
185                                                               target_container);
186
187                         memcpy(&key_addrs->v4addrs, &iph->saddr,
188                                sizeof(key_addrs->v4addrs));
189                         key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
190                 }
191
192                 if (ip_is_fragment(iph)) {
193                         key_control->flags |= FLOW_DIS_IS_FRAGMENT;
194
195                         if (iph->frag_off & htons(IP_OFFSET)) {
196                                 goto out_good;
197                         } else {
198                                 key_control->flags |= FLOW_DIS_FIRST_FRAG;
199                                 if (!(flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG))
200                                         goto out_good;
201                         }
202                 }
203
204                 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
205                         goto out_good;
206
207                 break;
208         }
209         case htons(ETH_P_IPV6): {
210                 const struct ipv6hdr *iph;
211                 struct ipv6hdr _iph;
212
213 ipv6:
214                 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
215                 if (!iph)
216                         goto out_bad;
217
218                 ip_proto = iph->nexthdr;
219                 nhoff += sizeof(struct ipv6hdr);
220
221                 if (dissector_uses_key(flow_dissector,
222                                        FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
223                         struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs;
224
225                         key_ipv6_addrs = skb_flow_dissector_target(flow_dissector,
226                                                                    FLOW_DISSECTOR_KEY_IPV6_ADDRS,
227                                                                    target_container);
228
229                         memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs));
230                         key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
231                 }
232
233                 if ((dissector_uses_key(flow_dissector,
234                                         FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
235                      (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
236                     ip6_flowlabel(iph)) {
237                         __be32 flow_label = ip6_flowlabel(iph);
238
239                         if (dissector_uses_key(flow_dissector,
240                                                FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
241                                 key_tags = skb_flow_dissector_target(flow_dissector,
242                                                                      FLOW_DISSECTOR_KEY_FLOW_LABEL,
243                                                                      target_container);
244                                 key_tags->flow_label = ntohl(flow_label);
245                         }
246                         if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)
247                                 goto out_good;
248                 }
249
250                 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
251                         goto out_good;
252
253                 break;
254         }
255         case htons(ETH_P_8021AD):
256         case htons(ETH_P_8021Q): {
257                 const struct vlan_hdr *vlan;
258                 struct vlan_hdr _vlan;
259
260                 vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan), data, hlen, &_vlan);
261                 if (!vlan)
262                         goto out_bad;
263
264                 if (dissector_uses_key(flow_dissector,
265                                        FLOW_DISSECTOR_KEY_VLANID)) {
266                         key_tags = skb_flow_dissector_target(flow_dissector,
267                                                              FLOW_DISSECTOR_KEY_VLANID,
268                                                              target_container);
269
270                         key_tags->vlan_id = skb_vlan_tag_get_id(skb);
271                 }
272
273                 proto = vlan->h_vlan_encapsulated_proto;
274                 nhoff += sizeof(*vlan);
275                 goto again;
276         }
277         case htons(ETH_P_PPP_SES): {
278                 struct {
279                         struct pppoe_hdr hdr;
280                         __be16 proto;
281                 } *hdr, _hdr;
282                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
283                 if (!hdr)
284                         goto out_bad;
285                 proto = hdr->proto;
286                 nhoff += PPPOE_SES_HLEN;
287                 switch (proto) {
288                 case htons(PPP_IP):
289                         goto ip;
290                 case htons(PPP_IPV6):
291                         goto ipv6;
292                 default:
293                         goto out_bad;
294                 }
295         }
296         case htons(ETH_P_TIPC): {
297                 struct {
298                         __be32 pre[3];
299                         __be32 srcnode;
300                 } *hdr, _hdr;
301                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
302                 if (!hdr)
303                         goto out_bad;
304
305                 if (dissector_uses_key(flow_dissector,
306                                        FLOW_DISSECTOR_KEY_TIPC_ADDRS)) {
307                         key_addrs = skb_flow_dissector_target(flow_dissector,
308                                                               FLOW_DISSECTOR_KEY_TIPC_ADDRS,
309                                                               target_container);
310                         key_addrs->tipcaddrs.srcnode = hdr->srcnode;
311                         key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC_ADDRS;
312                 }
313                 goto out_good;
314         }
315
316         case htons(ETH_P_MPLS_UC):
317         case htons(ETH_P_MPLS_MC): {
318                 struct mpls_label *hdr, _hdr[2];
319 mpls:
320                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
321                                            hlen, &_hdr);
322                 if (!hdr)
323                         goto out_bad;
324
325                 if ((ntohl(hdr[0].entry) & MPLS_LS_LABEL_MASK) >>
326                      MPLS_LS_LABEL_SHIFT == MPLS_LABEL_ENTROPY) {
327                         if (dissector_uses_key(flow_dissector,
328                                                FLOW_DISSECTOR_KEY_MPLS_ENTROPY)) {
329                                 key_keyid = skb_flow_dissector_target(flow_dissector,
330                                                                       FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
331                                                                       target_container);
332                                 key_keyid->keyid = hdr[1].entry &
333                                         htonl(MPLS_LS_LABEL_MASK);
334                         }
335
336                         goto out_good;
337                 }
338
339                 goto out_good;
340         }
341
342         case htons(ETH_P_FCOE):
343                 if ((hlen - nhoff) < FCOE_HEADER_LEN)
344                         goto out_bad;
345
346                 nhoff += FCOE_HEADER_LEN;
347                 goto out_good;
348         default:
349                 goto out_bad;
350         }
351
352 ip_proto_again:
353         switch (ip_proto) {
354         case IPPROTO_GRE: {
355                 struct gre_hdr {
356                         __be16 flags;
357                         __be16 proto;
358                 } *hdr, _hdr;
359
360                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
361                 if (!hdr)
362                         goto out_bad;
363                 /*
364                  * Only look inside GRE if version zero and no
365                  * routing
366                  */
367                 if (hdr->flags & (GRE_VERSION | GRE_ROUTING))
368                         break;
369
370                 proto = hdr->proto;
371                 nhoff += 4;
372                 if (hdr->flags & GRE_CSUM)
373                         nhoff += 4;
374                 if (hdr->flags & GRE_KEY) {
375                         const __be32 *keyid;
376                         __be32 _keyid;
377
378                         keyid = __skb_header_pointer(skb, nhoff, sizeof(_keyid),
379                                                      data, hlen, &_keyid);
380
381                         if (!keyid)
382                                 goto out_bad;
383
384                         if (dissector_uses_key(flow_dissector,
385                                                FLOW_DISSECTOR_KEY_GRE_KEYID)) {
386                                 key_keyid = skb_flow_dissector_target(flow_dissector,
387                                                                       FLOW_DISSECTOR_KEY_GRE_KEYID,
388                                                                       target_container);
389                                 key_keyid->keyid = *keyid;
390                         }
391                         nhoff += 4;
392                 }
393                 if (hdr->flags & GRE_SEQ)
394                         nhoff += 4;
395                 if (proto == htons(ETH_P_TEB)) {
396                         const struct ethhdr *eth;
397                         struct ethhdr _eth;
398
399                         eth = __skb_header_pointer(skb, nhoff,
400                                                    sizeof(_eth),
401                                                    data, hlen, &_eth);
402                         if (!eth)
403                                 goto out_bad;
404                         proto = eth->h_proto;
405                         nhoff += sizeof(*eth);
406
407                         /* Cap headers that we access via pointers at the
408                          * end of the Ethernet header as our maximum alignment
409                          * at that point is only 2 bytes.
410                          */
411                         if (NET_IP_ALIGN)
412                                 hlen = nhoff;
413                 }
414
415                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
416                 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
417                         goto out_good;
418
419                 goto again;
420         }
421         case NEXTHDR_HOP:
422         case NEXTHDR_ROUTING:
423         case NEXTHDR_DEST: {
424                 u8 _opthdr[2], *opthdr;
425
426                 if (proto != htons(ETH_P_IPV6))
427                         break;
428
429                 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
430                                               data, hlen, &_opthdr);
431                 if (!opthdr)
432                         goto out_bad;
433
434                 ip_proto = opthdr[0];
435                 nhoff += (opthdr[1] + 1) << 3;
436
437                 goto ip_proto_again;
438         }
439         case NEXTHDR_FRAGMENT: {
440                 struct frag_hdr _fh, *fh;
441
442                 if (proto != htons(ETH_P_IPV6))
443                         break;
444
445                 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
446                                           data, hlen, &_fh);
447
448                 if (!fh)
449                         goto out_bad;
450
451                 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
452
453                 nhoff += sizeof(_fh);
454                 ip_proto = fh->nexthdr;
455
456                 if (!(fh->frag_off & htons(IP6_OFFSET))) {
457                         key_control->flags |= FLOW_DIS_FIRST_FRAG;
458                         if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG)
459                                 goto ip_proto_again;
460                 }
461                 goto out_good;
462         }
463         case IPPROTO_IPIP:
464                 proto = htons(ETH_P_IP);
465
466                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
467                 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
468                         goto out_good;
469
470                 goto ip;
471         case IPPROTO_IPV6:
472                 proto = htons(ETH_P_IPV6);
473
474                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
475                 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
476                         goto out_good;
477
478                 goto ipv6;
479         case IPPROTO_MPLS:
480                 proto = htons(ETH_P_MPLS_UC);
481                 goto mpls;
482         default:
483                 break;
484         }
485
486         if (dissector_uses_key(flow_dissector,
487                                FLOW_DISSECTOR_KEY_PORTS)) {
488                 key_ports = skb_flow_dissector_target(flow_dissector,
489                                                       FLOW_DISSECTOR_KEY_PORTS,
490                                                       target_container);
491                 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
492                                                         data, hlen);
493         }
494
495 out_good:
496         ret = true;
497
498 out_bad:
499         key_basic->n_proto = proto;
500         key_basic->ip_proto = ip_proto;
501         key_control->thoff = (u16)nhoff;
502
503         return ret;
504 }
505 EXPORT_SYMBOL(__skb_flow_dissect);
506
507 static u32 hashrnd __read_mostly;
508 static __always_inline void __flow_hash_secret_init(void)
509 {
510         net_get_random_once(&hashrnd, sizeof(hashrnd));
511 }
512
513 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
514                                              u32 keyval)
515 {
516         return jhash2(words, length, keyval);
517 }
518
519 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
520 {
521         const void *p = flow;
522
523         BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
524         return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
525 }
526
527 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
528 {
529         size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
530         BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
531         BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
532                      sizeof(*flow) - sizeof(flow->addrs));
533
534         switch (flow->control.addr_type) {
535         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
536                 diff -= sizeof(flow->addrs.v4addrs);
537                 break;
538         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
539                 diff -= sizeof(flow->addrs.v6addrs);
540                 break;
541         case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
542                 diff -= sizeof(flow->addrs.tipcaddrs);
543                 break;
544         }
545         return (sizeof(*flow) - diff) / sizeof(u32);
546 }
547
548 __be32 flow_get_u32_src(const struct flow_keys *flow)
549 {
550         switch (flow->control.addr_type) {
551         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
552                 return flow->addrs.v4addrs.src;
553         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
554                 return (__force __be32)ipv6_addr_hash(
555                         &flow->addrs.v6addrs.src);
556         case FLOW_DISSECTOR_KEY_TIPC_ADDRS:
557                 return flow->addrs.tipcaddrs.srcnode;
558         default:
559                 return 0;
560         }
561 }
562 EXPORT_SYMBOL(flow_get_u32_src);
563
564 __be32 flow_get_u32_dst(const struct flow_keys *flow)
565 {
566         switch (flow->control.addr_type) {
567         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
568                 return flow->addrs.v4addrs.dst;
569         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
570                 return (__force __be32)ipv6_addr_hash(
571                         &flow->addrs.v6addrs.dst);
572         default:
573                 return 0;
574         }
575 }
576 EXPORT_SYMBOL(flow_get_u32_dst);
577
578 static inline void __flow_hash_consistentify(struct flow_keys *keys)
579 {
580         int addr_diff, i;
581
582         switch (keys->control.addr_type) {
583         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
584                 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
585                             (__force u32)keys->addrs.v4addrs.src;
586                 if ((addr_diff < 0) ||
587                     (addr_diff == 0 &&
588                      ((__force u16)keys->ports.dst <
589                       (__force u16)keys->ports.src))) {
590                         swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
591                         swap(keys->ports.src, keys->ports.dst);
592                 }
593                 break;
594         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
595                 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
596                                    &keys->addrs.v6addrs.src,
597                                    sizeof(keys->addrs.v6addrs.dst));
598                 if ((addr_diff < 0) ||
599                     (addr_diff == 0 &&
600                      ((__force u16)keys->ports.dst <
601                       (__force u16)keys->ports.src))) {
602                         for (i = 0; i < 4; i++)
603                                 swap(keys->addrs.v6addrs.src.s6_addr32[i],
604                                      keys->addrs.v6addrs.dst.s6_addr32[i]);
605                         swap(keys->ports.src, keys->ports.dst);
606                 }
607                 break;
608         }
609 }
610
611 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
612 {
613         u32 hash;
614
615         __flow_hash_consistentify(keys);
616
617         hash = __flow_hash_words(flow_keys_hash_start(keys),
618                                  flow_keys_hash_length(keys), keyval);
619         if (!hash)
620                 hash = 1;
621
622         return hash;
623 }
624
625 u32 flow_hash_from_keys(struct flow_keys *keys)
626 {
627         __flow_hash_secret_init();
628         return __flow_hash_from_keys(keys, hashrnd);
629 }
630 EXPORT_SYMBOL(flow_hash_from_keys);
631
632 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
633                                   struct flow_keys *keys, u32 keyval)
634 {
635         skb_flow_dissect_flow_keys(skb, keys,
636                                    FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
637
638         return __flow_hash_from_keys(keys, keyval);
639 }
640
641 struct _flow_keys_digest_data {
642         __be16  n_proto;
643         u8      ip_proto;
644         u8      padding;
645         __be32  ports;
646         __be32  src;
647         __be32  dst;
648 };
649
650 void make_flow_keys_digest(struct flow_keys_digest *digest,
651                            const struct flow_keys *flow)
652 {
653         struct _flow_keys_digest_data *data =
654             (struct _flow_keys_digest_data *)digest;
655
656         BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
657
658         memset(digest, 0, sizeof(*digest));
659
660         data->n_proto = flow->basic.n_proto;
661         data->ip_proto = flow->basic.ip_proto;
662         data->ports = flow->ports.ports;
663         data->src = flow->addrs.v4addrs.src;
664         data->dst = flow->addrs.v4addrs.dst;
665 }
666 EXPORT_SYMBOL(make_flow_keys_digest);
667
668 /**
669  * __skb_get_hash: calculate a flow hash
670  * @skb: sk_buff to calculate flow hash from
671  *
672  * This function calculates a flow hash based on src/dst addresses
673  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
674  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
675  * if hash is a canonical 4-tuple hash over transport ports.
676  */
677 void __skb_get_hash(struct sk_buff *skb)
678 {
679         struct flow_keys keys;
680
681         __flow_hash_secret_init();
682
683         __skb_set_sw_hash(skb, ___skb_get_hash(skb, &keys, hashrnd),
684                           flow_keys_have_l4(&keys));
685 }
686 EXPORT_SYMBOL(__skb_get_hash);
687
688 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
689 {
690         struct flow_keys keys;
691
692         return ___skb_get_hash(skb, &keys, perturb);
693 }
694 EXPORT_SYMBOL(skb_get_hash_perturb);
695
696 __u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
697 {
698         struct flow_keys keys;
699
700         memset(&keys, 0, sizeof(keys));
701
702         memcpy(&keys.addrs.v6addrs.src, &fl6->saddr,
703                sizeof(keys.addrs.v6addrs.src));
704         memcpy(&keys.addrs.v6addrs.dst, &fl6->daddr,
705                sizeof(keys.addrs.v6addrs.dst));
706         keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
707         keys.ports.src = fl6->fl6_sport;
708         keys.ports.dst = fl6->fl6_dport;
709         keys.keyid.keyid = fl6->fl6_gre_key;
710         keys.tags.flow_label = (__force u32)fl6->flowlabel;
711         keys.basic.ip_proto = fl6->flowi6_proto;
712
713         __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
714                           flow_keys_have_l4(&keys));
715
716         return skb->hash;
717 }
718 EXPORT_SYMBOL(__skb_get_hash_flowi6);
719
720 __u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
721 {
722         struct flow_keys keys;
723
724         memset(&keys, 0, sizeof(keys));
725
726         keys.addrs.v4addrs.src = fl4->saddr;
727         keys.addrs.v4addrs.dst = fl4->daddr;
728         keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
729         keys.ports.src = fl4->fl4_sport;
730         keys.ports.dst = fl4->fl4_dport;
731         keys.keyid.keyid = fl4->fl4_gre_key;
732         keys.basic.ip_proto = fl4->flowi4_proto;
733
734         __skb_set_sw_hash(skb, flow_hash_from_keys(&keys),
735                           flow_keys_have_l4(&keys));
736
737         return skb->hash;
738 }
739 EXPORT_SYMBOL(__skb_get_hash_flowi4);
740
741 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
742                    const struct flow_keys *keys, int hlen)
743 {
744         u32 poff = keys->control.thoff;
745
746         /* skip L4 headers for fragments after the first */
747         if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
748             !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
749                 return poff;
750
751         switch (keys->basic.ip_proto) {
752         case IPPROTO_TCP: {
753                 /* access doff as u8 to avoid unaligned access */
754                 const u8 *doff;
755                 u8 _doff;
756
757                 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
758                                             data, hlen, &_doff);
759                 if (!doff)
760                         return poff;
761
762                 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
763                 break;
764         }
765         case IPPROTO_UDP:
766         case IPPROTO_UDPLITE:
767                 poff += sizeof(struct udphdr);
768                 break;
769         /* For the rest, we do not really care about header
770          * extensions at this point for now.
771          */
772         case IPPROTO_ICMP:
773                 poff += sizeof(struct icmphdr);
774                 break;
775         case IPPROTO_ICMPV6:
776                 poff += sizeof(struct icmp6hdr);
777                 break;
778         case IPPROTO_IGMP:
779                 poff += sizeof(struct igmphdr);
780                 break;
781         case IPPROTO_DCCP:
782                 poff += sizeof(struct dccp_hdr);
783                 break;
784         case IPPROTO_SCTP:
785                 poff += sizeof(struct sctphdr);
786                 break;
787         }
788
789         return poff;
790 }
791
792 /**
793  * skb_get_poff - get the offset to the payload
794  * @skb: sk_buff to get the payload offset from
795  *
796  * The function will get the offset to the payload as far as it could
797  * be dissected.  The main user is currently BPF, so that we can dynamically
798  * truncate packets without needing to push actual payload to the user
799  * space and can analyze headers only, instead.
800  */
801 u32 skb_get_poff(const struct sk_buff *skb)
802 {
803         struct flow_keys keys;
804
805         if (!skb_flow_dissect_flow_keys(skb, &keys, 0))
806                 return 0;
807
808         return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
809 }
810
811 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
812 {
813         memset(keys, 0, sizeof(*keys));
814
815         memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
816             sizeof(keys->addrs.v6addrs.src));
817         memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
818             sizeof(keys->addrs.v6addrs.dst));
819         keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
820         keys->ports.src = fl6->fl6_sport;
821         keys->ports.dst = fl6->fl6_dport;
822         keys->keyid.keyid = fl6->fl6_gre_key;
823         keys->tags.flow_label = (__force u32)fl6->flowlabel;
824         keys->basic.ip_proto = fl6->flowi6_proto;
825
826         return flow_hash_from_keys(keys);
827 }
828 EXPORT_SYMBOL(__get_hash_from_flowi6);
829
830 __u32 __get_hash_from_flowi4(const struct flowi4 *fl4, struct flow_keys *keys)
831 {
832         memset(keys, 0, sizeof(*keys));
833
834         keys->addrs.v4addrs.src = fl4->saddr;
835         keys->addrs.v4addrs.dst = fl4->daddr;
836         keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
837         keys->ports.src = fl4->fl4_sport;
838         keys->ports.dst = fl4->fl4_dport;
839         keys->keyid.keyid = fl4->fl4_gre_key;
840         keys->basic.ip_proto = fl4->flowi4_proto;
841
842         return flow_hash_from_keys(keys);
843 }
844 EXPORT_SYMBOL(__get_hash_from_flowi4);
845
846 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
847         {
848                 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
849                 .offset = offsetof(struct flow_keys, control),
850         },
851         {
852                 .key_id = FLOW_DISSECTOR_KEY_BASIC,
853                 .offset = offsetof(struct flow_keys, basic),
854         },
855         {
856                 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
857                 .offset = offsetof(struct flow_keys, addrs.v4addrs),
858         },
859         {
860                 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
861                 .offset = offsetof(struct flow_keys, addrs.v6addrs),
862         },
863         {
864                 .key_id = FLOW_DISSECTOR_KEY_TIPC_ADDRS,
865                 .offset = offsetof(struct flow_keys, addrs.tipcaddrs),
866         },
867         {
868                 .key_id = FLOW_DISSECTOR_KEY_PORTS,
869                 .offset = offsetof(struct flow_keys, ports),
870         },
871         {
872                 .key_id = FLOW_DISSECTOR_KEY_VLANID,
873                 .offset = offsetof(struct flow_keys, tags),
874         },
875         {
876                 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
877                 .offset = offsetof(struct flow_keys, tags),
878         },
879         {
880                 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
881                 .offset = offsetof(struct flow_keys, keyid),
882         },
883 };
884
885 static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
886         {
887                 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
888                 .offset = offsetof(struct flow_keys, control),
889         },
890         {
891                 .key_id = FLOW_DISSECTOR_KEY_BASIC,
892                 .offset = offsetof(struct flow_keys, basic),
893         },
894 };
895
896 struct flow_dissector flow_keys_dissector __read_mostly;
897 EXPORT_SYMBOL(flow_keys_dissector);
898
899 struct flow_dissector flow_keys_buf_dissector __read_mostly;
900
901 static int __init init_default_flow_dissectors(void)
902 {
903         skb_flow_dissector_init(&flow_keys_dissector,
904                                 flow_keys_dissector_keys,
905                                 ARRAY_SIZE(flow_keys_dissector_keys));
906         skb_flow_dissector_init(&flow_keys_buf_dissector,
907                                 flow_keys_buf_dissector_keys,
908                                 ARRAY_SIZE(flow_keys_buf_dissector_keys));
909         return 0;
910 }
911
912 late_initcall_sync(init_default_flow_dissectors);