net/core/flow_dissector.c

   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 <net/dsa.h>
   9 #include <net/dst_metadata.h>
  10 #include <net/ip.h>
  11 #include <net/ipv6.h>
  12 #include <net/gre.h>
  13 #include <net/pptp.h>
  14 #include <net/tipc.h>
  15 #include <linux/igmp.h>
  16 #include <linux/icmp.h>
  17 #include <linux/sctp.h>
  18 #include <linux/dccp.h>
  19 #include <linux/if_tunnel.h>
  20 #include <linux/if_pppox.h>
  21 #include <linux/ppp_defs.h>
  22 #include <linux/stddef.h>
  23 #include <linux/if_ether.h>
  24 #include <linux/mpls.h>
  25 #include <linux/tcp.h>
  26 #include <net/flow_dissector.h>
  27 #include <scsi/fc/fc_fcoe.h>
  28 #include <uapi/linux/batadv_packet.h>
  29 #include <linux/bpf.h>
  30
  31 static DEFINE_MUTEX(flow_dissector_mutex);
  32
  33 static void dissector_set_key(struct flow_dissector *flow_dissector,
  34                               enum flow_dissector_key_id key_id)
  35 {
  36         flow_dissector->used_keys |= (1 << key_id);
  37 }
  38
  39 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
  40                              const struct flow_dissector_key *key,
  41                              unsigned int key_count)
  42 {
  43         unsigned int i;
  44
  45         memset(flow_dissector, 0, sizeof(*flow_dissector));
  46
  47         for (i = 0; i < key_count; i++, key++) {
  48                 /* User should make sure that every key target offset is withing
  49                  * boundaries of unsigned short.
  50                  */
  51                 BUG_ON(key->offset > USHRT_MAX);
  52                 BUG_ON(dissector_uses_key(flow_dissector,
  53                                           key->key_id));
  54
  55                 dissector_set_key(flow_dissector, key->key_id);
  56                 flow_dissector->offset[key->key_id] = key->offset;
  57         }
  58
  59         /* Ensure that the dissector always includes control and basic key.
  60          * That way we are able to avoid handling lack of these in fast path.
  61          */
  62         BUG_ON(!dissector_uses_key(flow_dissector,
  63                                    FLOW_DISSECTOR_KEY_CONTROL));
  64         BUG_ON(!dissector_uses_key(flow_dissector,
  65                                    FLOW_DISSECTOR_KEY_BASIC));
  66 }
  67 EXPORT_SYMBOL(skb_flow_dissector_init);
  68
  69 int skb_flow_dissector_prog_query(const union bpf_attr *attr,
  70                                   union bpf_attr __user *uattr)
  71 {
  72         __u32 __user *prog_ids = u64_to_user_ptr(attr->query.prog_ids);
  73         u32 prog_id, prog_cnt = 0, flags = 0;
  74         struct bpf_prog *attached;
  75         struct net *net;
  76
  77         if (attr->query.query_flags)
  78                 return -EINVAL;
  79
  80         net = get_net_ns_by_fd(attr->query.target_fd);
  81         if (IS_ERR(net))
  82                 return PTR_ERR(net);
  83
  84         rcu_read_lock();
  85         attached = rcu_dereference(net->flow_dissector_prog);
  86         if (attached) {
  87                 prog_cnt = 1;
  88                 prog_id = attached->aux->id;
  89         }
  90         rcu_read_unlock();
  91
  92         put_net(net);
  93
  94         if (copy_to_user(&uattr->query.attach_flags, &flags, sizeof(flags)))
  95                 return -EFAULT;
  96         if (copy_to_user(&uattr->query.prog_cnt, &prog_cnt, sizeof(prog_cnt)))
  97                 return -EFAULT;
  98
  99         if (!attr->query.prog_cnt || !prog_ids || !prog_cnt)
 100                 return 0;
 101
 102         if (copy_to_user(prog_ids, &prog_id, sizeof(u32)))
 103                 return -EFAULT;
 104
 105         return 0;
 106 }
 107
 108 int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
 109                                        struct bpf_prog *prog)
 110 {
 111         struct bpf_prog *attached;
 112         struct net *net;
 113
 114         net = current->nsproxy->net_ns;
 115         mutex_lock(&flow_dissector_mutex);
 116         attached = rcu_dereference_protected(net->flow_dissector_prog,
 117                                              lockdep_is_held(&flow_dissector_mutex));
 118         if (attached) {
 119                 /* Only one BPF program can be attached at a time */
 120                 mutex_unlock(&flow_dissector_mutex);
 121                 return -EEXIST;
 122         }
 123         rcu_assign_pointer(net->flow_dissector_prog, prog);
 124         mutex_unlock(&flow_dissector_mutex);
 125         return 0;
 126 }
 127
 128 int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
 129 {
 130         struct bpf_prog *attached;
 131         struct net *net;
 132
 133         net = current->nsproxy->net_ns;
 134         mutex_lock(&flow_dissector_mutex);
 135         attached = rcu_dereference_protected(net->flow_dissector_prog,
 136                                              lockdep_is_held(&flow_dissector_mutex));
 137         if (!attached) {
 138                 mutex_unlock(&flow_dissector_mutex);
 139                 return -ENOENT;
 140         }
 141         bpf_prog_put(attached);
 142         RCU_INIT_POINTER(net->flow_dissector_prog, NULL);
 143         mutex_unlock(&flow_dissector_mutex);
 144         return 0;
 145 }
 146 /**
 147  * skb_flow_get_be16 - extract be16 entity
 148  * @skb: sk_buff to extract from
 149  * @poff: offset to extract at
 150  * @data: raw buffer pointer to the packet
 151  * @hlen: packet header length
 152  *
 153  * The function will try to retrieve a be32 entity at
 154  * offset poff
 155  */
 156 static __be16 skb_flow_get_be16(const struct sk_buff *skb, int poff,
 157                                 void *data, int hlen)
 158 {
 159         __be16 *u, _u;
 160
 161         u = __skb_header_pointer(skb, poff, sizeof(_u), data, hlen, &_u);
 162         if (u)
 163                 return *u;
 164
 165         return 0;
 166 }
 167
 168 /**
 169  * __skb_flow_get_ports - extract the upper layer ports and return them
 170  * @skb: sk_buff to extract the ports from
 171  * @thoff: transport header offset
 172  * @ip_proto: protocol for which to get port offset
 173  * @data: raw buffer pointer to the packet, if NULL use skb->data
 174  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
 175  *
 176  * The function will try to retrieve the ports at offset thoff + poff where poff
 177  * is the protocol port offset returned from proto_ports_offset
 178  */
 179 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
 180                             void *data, int hlen)
 181 {
 182         int poff = proto_ports_offset(ip_proto);
 183
 184         if (!data) {
 185                 data = skb->data;
 186                 hlen = skb_headlen(skb);
 187         }
 188
 189         if (poff >= 0) {
 190                 __be32 *ports, _ports;
 191
 192                 ports = __skb_header_pointer(skb, thoff + poff,
 193                                              sizeof(_ports), data, hlen, &_ports);
 194                 if (ports)
 195                         return *ports;
 196         }
 197
 198         return 0;
 199 }
 200 EXPORT_SYMBOL(__skb_flow_get_ports);
 201
 202 static void
 203 skb_flow_dissect_set_enc_addr_type(enum flow_dissector_key_id type,
 204                                    struct flow_dissector *flow_dissector,
 205                                    void *target_container)
 206 {
 207         struct flow_dissector_key_control *ctrl;
 208
 209         if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_CONTROL))
 210                 return;
 211
 212         ctrl = skb_flow_dissector_target(flow_dissector,
 213                                          FLOW_DISSECTOR_KEY_ENC_CONTROL,
 214                                          target_container);
 215         ctrl->addr_type = type;
 216 }
 217
 218 void
 219 skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
 220                              struct flow_dissector *flow_dissector,
 221                              void *target_container)
 222 {
 223         struct ip_tunnel_info *info;
 224         struct ip_tunnel_key *key;
 225
 226         /* A quick check to see if there might be something to do. */
 227         if (!dissector_uses_key(flow_dissector,
 228                                 FLOW_DISSECTOR_KEY_ENC_KEYID) &&
 229             !dissector_uses_key(flow_dissector,
 230                                 FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS) &&
 231             !dissector_uses_key(flow_dissector,
 232                                 FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS) &&
 233             !dissector_uses_key(flow_dissector,
 234                                 FLOW_DISSECTOR_KEY_ENC_CONTROL) &&
 235             !dissector_uses_key(flow_dissector,
 236                                 FLOW_DISSECTOR_KEY_ENC_PORTS) &&
 237             !dissector_uses_key(flow_dissector,
 238                                 FLOW_DISSECTOR_KEY_ENC_IP) &&
 239             !dissector_uses_key(flow_dissector,
 240                                 FLOW_DISSECTOR_KEY_ENC_OPTS))
 241                 return;
 242
 243         info = skb_tunnel_info(skb);
 244         if (!info)
 245                 return;
 246
 247         key = &info->key;
 248
 249         switch (ip_tunnel_info_af(info)) {
 250         case AF_INET:
 251                 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV4_ADDRS,
 252                                                    flow_dissector,
 253                                                    target_container);
 254                 if (dissector_uses_key(flow_dissector,
 255                                        FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS)) {
 256                         struct flow_dissector_key_ipv4_addrs *ipv4;
 257
 258                         ipv4 = skb_flow_dissector_target(flow_dissector,
 259                                                          FLOW_DISSECTOR_KEY_ENC_IPV4_ADDRS,
 260                                                          target_container);
 261                         ipv4->src = key->u.ipv4.src;
 262                         ipv4->dst = key->u.ipv4.dst;
 263                 }
 264                 break;
 265         case AF_INET6:
 266                 skb_flow_dissect_set_enc_addr_type(FLOW_DISSECTOR_KEY_IPV6_ADDRS,
 267                                                    flow_dissector,
 268                                                    target_container);
 269                 if (dissector_uses_key(flow_dissector,
 270                                        FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS)) {
 271                         struct flow_dissector_key_ipv6_addrs *ipv6;
 272
 273                         ipv6 = skb_flow_dissector_target(flow_dissector,
 274                                                          FLOW_DISSECTOR_KEY_ENC_IPV6_ADDRS,
 275                                                          target_container);
 276                         ipv6->src = key->u.ipv6.src;
 277                         ipv6->dst = key->u.ipv6.dst;
 278                 }
 279                 break;
 280         }
 281
 282         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_KEYID)) {
 283                 struct flow_dissector_key_keyid *keyid;
 284
 285                 keyid = skb_flow_dissector_target(flow_dissector,
 286                                                   FLOW_DISSECTOR_KEY_ENC_KEYID,
 287                                                   target_container);
 288                 keyid->keyid = tunnel_id_to_key32(key->tun_id);
 289         }
 290
 291         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_PORTS)) {
 292                 struct flow_dissector_key_ports *tp;
 293
 294                 tp = skb_flow_dissector_target(flow_dissector,
 295                                                FLOW_DISSECTOR_KEY_ENC_PORTS,
 296                                                target_container);
 297                 tp->src = key->tp_src;
 298                 tp->dst = key->tp_dst;
 299         }
 300
 301         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_IP)) {
 302                 struct flow_dissector_key_ip *ip;
 303
 304                 ip = skb_flow_dissector_target(flow_dissector,
 305                                                FLOW_DISSECTOR_KEY_ENC_IP,
 306                                                target_container);
 307                 ip->tos = key->tos;
 308                 ip->ttl = key->ttl;
 309         }
 310
 311         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ENC_OPTS)) {
 312                 struct flow_dissector_key_enc_opts *enc_opt;
 313
 314                 enc_opt = skb_flow_dissector_target(flow_dissector,
 315                                                     FLOW_DISSECTOR_KEY_ENC_OPTS,
 316                                                     target_container);
 317
 318                 if (info->options_len) {
 319                         enc_opt->len = info->options_len;
 320                         ip_tunnel_info_opts_get(enc_opt->data, info);
 321                         enc_opt->dst_opt_type = info->key.tun_flags &
 322                                                 TUNNEL_OPTIONS_PRESENT;
 323                 }
 324         }
 325 }
 326 EXPORT_SYMBOL(skb_flow_dissect_tunnel_info);
 327
 328 static enum flow_dissect_ret
 329 __skb_flow_dissect_mpls(const struct sk_buff *skb,
 330                         struct flow_dissector *flow_dissector,
 331                         void *target_container, void *data, int nhoff, int hlen)
 332 {
 333         struct flow_dissector_key_keyid *key_keyid;
 334         struct mpls_label *hdr, _hdr[2];
 335         u32 entry, label;
 336
 337         if (!dissector_uses_key(flow_dissector,
 338                                 FLOW_DISSECTOR_KEY_MPLS_ENTROPY) &&
 339             !dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS))
 340                 return FLOW_DISSECT_RET_OUT_GOOD;
 341
 342         hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data,
 343                                    hlen, &_hdr);
 344         if (!hdr)
 345                 return FLOW_DISSECT_RET_OUT_BAD;
 346
 347         entry = ntohl(hdr[0].entry);
 348         label = (entry & MPLS_LS_LABEL_MASK) >> MPLS_LS_LABEL_SHIFT;
 349
 350         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_MPLS)) {
 351                 struct flow_dissector_key_mpls *key_mpls;
 352
 353                 key_mpls = skb_flow_dissector_target(flow_dissector,
 354                                                      FLOW_DISSECTOR_KEY_MPLS,
 355                                                      target_container);
 356                 key_mpls->mpls_label = label;
 357                 key_mpls->mpls_ttl = (entry & MPLS_LS_TTL_MASK)
 358                                         >> MPLS_LS_TTL_SHIFT;
 359                 key_mpls->mpls_tc = (entry & MPLS_LS_TC_MASK)
 360                                         >> MPLS_LS_TC_SHIFT;
 361                 key_mpls->mpls_bos = (entry & MPLS_LS_S_MASK)
 362                                         >> MPLS_LS_S_SHIFT;
 363         }
 364
 365         if (label == MPLS_LABEL_ENTROPY) {
 366                 key_keyid = skb_flow_dissector_target(flow_dissector,
 367                                                       FLOW_DISSECTOR_KEY_MPLS_ENTROPY,
 368                                                       target_container);
 369                 key_keyid->keyid = hdr[1].entry & htonl(MPLS_LS_LABEL_MASK);
 370         }
 371         return FLOW_DISSECT_RET_OUT_GOOD;
 372 }
 373
 374 static enum flow_dissect_ret
 375 __skb_flow_dissect_arp(const struct sk_buff *skb,
 376                        struct flow_dissector *flow_dissector,
 377                        void *target_container, void *data, int nhoff, int hlen)
 378 {
 379         struct flow_dissector_key_arp *key_arp;
 380         struct {
 381                 unsigned char ar_sha[ETH_ALEN];
 382                 unsigned char ar_sip[4];
 383                 unsigned char ar_tha[ETH_ALEN];
 384                 unsigned char ar_tip[4];
 385         } *arp_eth, _arp_eth;
 386         const struct arphdr *arp;
 387         struct arphdr _arp;
 388
 389         if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_ARP))
 390                 return FLOW_DISSECT_RET_OUT_GOOD;
 391
 392         arp = __skb_header_pointer(skb, nhoff, sizeof(_arp), data,
 393                                    hlen, &_arp);
 394         if (!arp)
 395                 return FLOW_DISSECT_RET_OUT_BAD;
 396
 397         if (arp->ar_hrd != htons(ARPHRD_ETHER) ||
 398             arp->ar_pro != htons(ETH_P_IP) ||
 399             arp->ar_hln != ETH_ALEN ||
 400             arp->ar_pln != 4 ||
 401             (arp->ar_op != htons(ARPOP_REPLY) &&
 402              arp->ar_op != htons(ARPOP_REQUEST)))
 403                 return FLOW_DISSECT_RET_OUT_BAD;
 404
 405         arp_eth = __skb_header_pointer(skb, nhoff + sizeof(_arp),
 406                                        sizeof(_arp_eth), data,
 407                                        hlen, &_arp_eth);
 408         if (!arp_eth)
 409                 return FLOW_DISSECT_RET_OUT_BAD;
 410
 411         key_arp = skb_flow_dissector_target(flow_dissector,
 412                                             FLOW_DISSECTOR_KEY_ARP,
 413                                             target_container);
 414
 415         memcpy(&key_arp->sip, arp_eth->ar_sip, sizeof(key_arp->sip));
 416         memcpy(&key_arp->tip, arp_eth->ar_tip, sizeof(key_arp->tip));
 417
 418         /* Only store the lower byte of the opcode;
 419          * this covers ARPOP_REPLY and ARPOP_REQUEST.
 420          */
 421         key_arp->op = ntohs(arp->ar_op) & 0xff;
 422
 423         ether_addr_copy(key_arp->sha, arp_eth->ar_sha);
 424         ether_addr_copy(key_arp->tha, arp_eth->ar_tha);
 425
 426         return FLOW_DISSECT_RET_OUT_GOOD;
 427 }
 428
 429 static enum flow_dissect_ret
 430 __skb_flow_dissect_gre(const struct sk_buff *skb,
 431                        struct flow_dissector_key_control *key_control,
 432                        struct flow_dissector *flow_dissector,
 433                        void *target_container, void *data,
 434                        __be16 *p_proto, int *p_nhoff, int *p_hlen,
 435                        unsigned int flags)
 436 {
 437         struct flow_dissector_key_keyid *key_keyid;
 438         struct gre_base_hdr *hdr, _hdr;
 439         int offset = 0;
 440         u16 gre_ver;
 441
 442         hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr),
 443                                    data, *p_hlen, &_hdr);
 444         if (!hdr)
 445                 return FLOW_DISSECT_RET_OUT_BAD;
 446
 447         /* Only look inside GRE without routing */
 448         if (hdr->flags & GRE_ROUTING)
 449                 return FLOW_DISSECT_RET_OUT_GOOD;
 450
 451         /* Only look inside GRE for version 0 and 1 */
 452         gre_ver = ntohs(hdr->flags & GRE_VERSION);
 453         if (gre_ver > 1)
 454                 return FLOW_DISSECT_RET_OUT_GOOD;
 455
 456         *p_proto = hdr->protocol;
 457         if (gre_ver) {
 458                 /* Version1 must be PPTP, and check the flags */
 459                 if (!(*p_proto == GRE_PROTO_PPP && (hdr->flags & GRE_KEY)))
 460                         return FLOW_DISSECT_RET_OUT_GOOD;
 461         }
 462
 463         offset += sizeof(struct gre_base_hdr);
 464
 465         if (hdr->flags & GRE_CSUM)
 466                 offset += FIELD_SIZEOF(struct gre_full_hdr, csum) +
 467                           FIELD_SIZEOF(struct gre_full_hdr, reserved1);
 468
 469         if (hdr->flags & GRE_KEY) {
 470                 const __be32 *keyid;
 471                 __be32 _keyid;
 472
 473                 keyid = __skb_header_pointer(skb, *p_nhoff + offset,
 474                                              sizeof(_keyid),
 475                                              data, *p_hlen, &_keyid);
 476                 if (!keyid)
 477                         return FLOW_DISSECT_RET_OUT_BAD;
 478
 479                 if (dissector_uses_key(flow_dissector,
 480                                        FLOW_DISSECTOR_KEY_GRE_KEYID)) {
 481                         key_keyid = skb_flow_dissector_target(flow_dissector,
 482                                                               FLOW_DISSECTOR_KEY_GRE_KEYID,
 483                                                               target_container);
 484                         if (gre_ver == 0)
 485                                 key_keyid->keyid = *keyid;
 486                         else
 487                                 key_keyid->keyid = *keyid & GRE_PPTP_KEY_MASK;
 488                 }
 489                 offset += FIELD_SIZEOF(struct gre_full_hdr, key);
 490         }
 491
 492         if (hdr->flags & GRE_SEQ)
 493                 offset += FIELD_SIZEOF(struct pptp_gre_header, seq);
 494
 495         if (gre_ver == 0) {
 496                 if (*p_proto == htons(ETH_P_TEB)) {
 497                         const struct ethhdr *eth;
 498                         struct ethhdr _eth;
 499
 500                         eth = __skb_header_pointer(skb, *p_nhoff + offset,
 501                                                    sizeof(_eth),
 502                                                    data, *p_hlen, &_eth);
 503                         if (!eth)
 504                                 return FLOW_DISSECT_RET_OUT_BAD;
 505                         *p_proto = eth->h_proto;
 506                         offset += sizeof(*eth);
 507
 508                         /* Cap headers that we access via pointers at the
 509                          * end of the Ethernet header as our maximum alignment
 510                          * at that point is only 2 bytes.
 511                          */
 512                         if (NET_IP_ALIGN)
 513                                 *p_hlen = *p_nhoff + offset;
 514                 }
 515         } else { /* version 1, must be PPTP */
 516                 u8 _ppp_hdr[PPP_HDRLEN];
 517                 u8 *ppp_hdr;
 518
 519                 if (hdr->flags & GRE_ACK)
 520                         offset += FIELD_SIZEOF(struct pptp_gre_header, ack);
 521
 522                 ppp_hdr = __skb_header_pointer(skb, *p_nhoff + offset,
 523                                                sizeof(_ppp_hdr),
 524                                                data, *p_hlen, _ppp_hdr);
 525                 if (!ppp_hdr)
 526                         return FLOW_DISSECT_RET_OUT_BAD;
 527
 528                 switch (PPP_PROTOCOL(ppp_hdr)) {
 529                 case PPP_IP:
 530                         *p_proto = htons(ETH_P_IP);
 531                         break;
 532                 case PPP_IPV6:
 533                         *p_proto = htons(ETH_P_IPV6);
 534                         break;
 535                 default:
 536                         /* Could probably catch some more like MPLS */
 537                         break;
 538                 }
 539
 540                 offset += PPP_HDRLEN;
 541         }
 542
 543         *p_nhoff += offset;
 544         key_control->flags |= FLOW_DIS_ENCAPSULATION;
 545         if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
 546                 return FLOW_DISSECT_RET_OUT_GOOD;
 547
 548         return FLOW_DISSECT_RET_PROTO_AGAIN;
 549 }
 550
 551 /**
 552  * __skb_flow_dissect_batadv() - dissect batman-adv header
 553  * @skb: sk_buff to with the batman-adv header
 554  * @key_control: flow dissectors control key
 555  * @data: raw buffer pointer to the packet, if NULL use skb->data
 556  * @p_proto: pointer used to update the protocol to process next
 557  * @p_nhoff: pointer used to update inner network header offset
 558  * @hlen: packet header length
 559  * @flags: any combination of FLOW_DISSECTOR_F_*
 560  *
 561  * ETH_P_BATMAN packets are tried to be dissected. Only
 562  * &struct batadv_unicast packets are actually processed because they contain an
 563  * inner ethernet header and are usually followed by actual network header. This
 564  * allows the flow dissector to continue processing the packet.
 565  *
 566  * Return: FLOW_DISSECT_RET_PROTO_AGAIN when &struct batadv_unicast was found,
 567  *  FLOW_DISSECT_RET_OUT_GOOD when dissector should stop after encapsulation,
 568  *  otherwise FLOW_DISSECT_RET_OUT_BAD
 569  */
 570 static enum flow_dissect_ret
 571 __skb_flow_dissect_batadv(const struct sk_buff *skb,
 572                           struct flow_dissector_key_control *key_control,
 573                           void *data, __be16 *p_proto, int *p_nhoff, int hlen,
 574                           unsigned int flags)
 575 {
 576         struct {
 577                 struct batadv_unicast_packet batadv_unicast;
 578                 struct ethhdr eth;
 579         } *hdr, _hdr;
 580
 581         hdr = __skb_header_pointer(skb, *p_nhoff, sizeof(_hdr), data, hlen,
 582                                    &_hdr);
 583         if (!hdr)
 584                 return FLOW_DISSECT_RET_OUT_BAD;
 585
 586         if (hdr->batadv_unicast.version != BATADV_COMPAT_VERSION)
 587                 return FLOW_DISSECT_RET_OUT_BAD;
 588
 589         if (hdr->batadv_unicast.packet_type != BATADV_UNICAST)
 590                 return FLOW_DISSECT_RET_OUT_BAD;
 591
 592         *p_proto = hdr->eth.h_proto;
 593         *p_nhoff += sizeof(*hdr);
 594
 595         key_control->flags |= FLOW_DIS_ENCAPSULATION;
 596         if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP)
 597                 return FLOW_DISSECT_RET_OUT_GOOD;
 598
 599         return FLOW_DISSECT_RET_PROTO_AGAIN;
 600 }
 601
 602 static void
 603 __skb_flow_dissect_tcp(const struct sk_buff *skb,
 604                        struct flow_dissector *flow_dissector,
 605                        void *target_container, void *data, int thoff, int hlen)
 606 {
 607         struct flow_dissector_key_tcp *key_tcp;
 608         struct tcphdr *th, _th;
 609
 610         if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_TCP))
 611                 return;
 612
 613         th = __skb_header_pointer(skb, thoff, sizeof(_th), data, hlen, &_th);
 614         if (!th)
 615                 return;
 616
 617         if (unlikely(__tcp_hdrlen(th) < sizeof(_th)))
 618                 return;
 619
 620         key_tcp = skb_flow_dissector_target(flow_dissector,
 621                                             FLOW_DISSECTOR_KEY_TCP,
 622                                             target_container);
 623         key_tcp->flags = (*(__be16 *) &tcp_flag_word(th) & htons(0x0FFF));
 624 }
 625
 626 static void
 627 __skb_flow_dissect_ipv4(const struct sk_buff *skb,
 628                         struct flow_dissector *flow_dissector,
 629                         void *target_container, void *data, const struct iphdr *iph)
 630 {
 631         struct flow_dissector_key_ip *key_ip;
 632
 633         if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
 634                 return;
 635
 636         key_ip = skb_flow_dissector_target(flow_dissector,
 637                                            FLOW_DISSECTOR_KEY_IP,
 638                                            target_container);
 639         key_ip->tos = iph->tos;
 640         key_ip->ttl = iph->ttl;
 641 }
 642
 643 static void
 644 __skb_flow_dissect_ipv6(const struct sk_buff *skb,
 645                         struct flow_dissector *flow_dissector,
 646                         void *target_container, void *data, const struct ipv6hdr *iph)
 647 {
 648         struct flow_dissector_key_ip *key_ip;
 649
 650         if (!dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IP))
 651                 return;
 652
 653         key_ip = skb_flow_dissector_target(flow_dissector,
 654                                            FLOW_DISSECTOR_KEY_IP,
 655                                            target_container);
 656         key_ip->tos = ipv6_get_dsfield(iph);
 657         key_ip->ttl = iph->hop_limit;
 658 }
 659
 660 /* Maximum number of protocol headers that can be parsed in
 661  * __skb_flow_dissect
 662  */
 663 #define MAX_FLOW_DISSECT_HDRS   15
 664
 665 static bool skb_flow_dissect_allowed(int *num_hdrs)
 666 {
 667         ++*num_hdrs;
 668
 669         return (*num_hdrs <= MAX_FLOW_DISSECT_HDRS);
 670 }
 671
 672 static void __skb_flow_bpf_to_target(const struct bpf_flow_keys *flow_keys,
 673                                      struct flow_dissector *flow_dissector,
 674                                      void *target_container)
 675 {
 676         struct flow_dissector_key_control *key_control;
 677         struct flow_dissector_key_basic *key_basic;
 678         struct flow_dissector_key_addrs *key_addrs;
 679         struct flow_dissector_key_ports *key_ports;
 680
 681         key_control = skb_flow_dissector_target(flow_dissector,
 682                                                 FLOW_DISSECTOR_KEY_CONTROL,
 683                                                 target_container);
 684         key_control->thoff = flow_keys->thoff;
 685         if (flow_keys->is_frag)
 686                 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
 687         if (flow_keys->is_first_frag)
 688                 key_control->flags |= FLOW_DIS_FIRST_FRAG;
 689         if (flow_keys->is_encap)
 690                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
 691
 692         key_basic = skb_flow_dissector_target(flow_dissector,
 693                                               FLOW_DISSECTOR_KEY_BASIC,
 694                                               target_container);
 695         key_basic->n_proto = flow_keys->n_proto;
 696         key_basic->ip_proto = flow_keys->ip_proto;
 697
 698         if (flow_keys->addr_proto == ETH_P_IP &&
 699             dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
 700                 key_addrs = skb_flow_dissector_target(flow_dissector,
 701                                                       FLOW_DISSECTOR_KEY_IPV4_ADDRS,
 702                                                       target_container);
 703                 key_addrs->v4addrs.src = flow_keys->ipv4_src;
 704                 key_addrs->v4addrs.dst = flow_keys->ipv4_dst;
 705                 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
 706         } else if (flow_keys->addr_proto == ETH_P_IPV6 &&
 707                    dissector_uses_key(flow_dissector,
 708                                       FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
 709                 key_addrs = skb_flow_dissector_target(flow_dissector,
 710                                                       FLOW_DISSECTOR_KEY_IPV6_ADDRS,
 711                                                       target_container);
 712                 memcpy(&key_addrs->v6addrs, &flow_keys->ipv6_src,
 713                        sizeof(key_addrs->v6addrs));
 714                 key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
 715         }
 716
 717         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS)) {
 718                 key_ports = skb_flow_dissector_target(flow_dissector,
 719                                                       FLOW_DISSECTOR_KEY_PORTS,
 720                                                       target_container);
 721                 key_ports->src = flow_keys->sport;
 722                 key_ports->dst = flow_keys->dport;
 723         }
 724 }
 725
 726 bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
 727                       __be16 proto, int nhoff, int hlen)
 728 {
 729         struct bpf_flow_keys *flow_keys = ctx->flow_keys;
 730         u32 result;
 731
 732         /* Pass parameters to the BPF program */
 733         memset(flow_keys, 0, sizeof(*flow_keys));
 734         flow_keys->n_proto = proto;
 735         flow_keys->nhoff = nhoff;
 736         flow_keys->thoff = flow_keys->nhoff;
 737
 738         preempt_disable();
 739         result = BPF_PROG_RUN(prog, ctx);
 740         preempt_enable();
 741
 742         flow_keys->nhoff = clamp_t(u16, flow_keys->nhoff, nhoff, hlen);
 743         flow_keys->thoff = clamp_t(u16, flow_keys->thoff,
 744                                    flow_keys->nhoff, hlen);
 745
 746         return result == BPF_OK;
 747 }
 748
 749 /**
 750  * __skb_flow_dissect - extract the flow_keys struct and return it
 751  * @net: associated network namespace, derived from @skb if NULL
 752  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
 753  * @flow_dissector: list of keys to dissect
 754  * @target_container: target structure to put dissected values into
 755  * @data: raw buffer pointer to the packet, if NULL use skb->data
 756  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
 757  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
 758  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
 759  * @flags: flags that control the dissection process, e.g.
 760  *         FLOW_DISSECTOR_F_STOP_AT_L3.
 761  *
 762  * The function will try to retrieve individual keys into target specified
 763  * by flow_dissector from either the skbuff or a raw buffer specified by the
 764  * rest parameters.
 765  *
 766  * Caller must take care of zeroing target container memory.
 767  */
 768 bool __skb_flow_dissect(const struct net *net,
 769                         const struct sk_buff *skb,
 770                         struct flow_dissector *flow_dissector,
 771                         void *target_container,
 772                         void *data, __be16 proto, int nhoff, int hlen,
 773                         unsigned int flags)
 774 {
 775         struct flow_dissector_key_control *key_control;
 776         struct flow_dissector_key_basic *key_basic;
 777         struct flow_dissector_key_addrs *key_addrs;
 778         struct flow_dissector_key_ports *key_ports;
 779         struct flow_dissector_key_icmp *key_icmp;
 780         struct flow_dissector_key_tags *key_tags;
 781         struct flow_dissector_key_vlan *key_vlan;
 782         struct bpf_prog *attached = NULL;
 783         enum flow_dissect_ret fdret;
 784         enum flow_dissector_key_id dissector_vlan = FLOW_DISSECTOR_KEY_MAX;
 785         int num_hdrs = 0;
 786         u8 ip_proto = 0;
 787         bool ret;
 788
 789         if (!data) {
 790                 data = skb->data;
 791                 proto = skb_vlan_tag_present(skb) ?
 792                          skb->vlan_proto : skb->protocol;
 793                 nhoff = skb_network_offset(skb);
 794                 hlen = skb_headlen(skb);
 795 #if IS_ENABLED(CONFIG_NET_DSA)
 796                 if (unlikely(skb->dev && netdev_uses_dsa(skb->dev))) {
 797                         const struct dsa_device_ops *ops;
 798                         int offset;
 799
 800                         ops = skb->dev->dsa_ptr->tag_ops;
 801                         if (ops->flow_dissect &&
 802                             !ops->flow_dissect(skb, &proto, &offset)) {
 803                                 hlen -= offset;
 804                                 nhoff += offset;
 805                         }
 806                 }
 807 #endif
 808         }
 809
 810         /* It is ensured by skb_flow_dissector_init() that control key will
 811          * be always present.
 812          */
 813         key_control = skb_flow_dissector_target(flow_dissector,
 814                                                 FLOW_DISSECTOR_KEY_CONTROL,
 815                                                 target_container);
 816
 817         /* It is ensured by skb_flow_dissector_init() that basic key will
 818          * be always present.
 819          */
 820         key_basic = skb_flow_dissector_target(flow_dissector,
 821                                               FLOW_DISSECTOR_KEY_BASIC,
 822                                               target_container);
 823
 824         if (skb) {
 825                 if (!net) {
 826                         if (skb->dev)
 827                                 net = dev_net(skb->dev);
 828                         else if (skb->sk)
 829                                 net = sock_net(skb->sk);
 830                 }
 831         }
 832
 833         WARN_ON_ONCE(!net);
 834         if (net) {
 835                 rcu_read_lock();
 836                 attached = rcu_dereference(net->flow_dissector_prog);
 837
 838                 if (attached) {
 839                         struct bpf_flow_keys flow_keys;
 840                         struct bpf_flow_dissector ctx = {
 841                                 .flow_keys = &flow_keys,
 842                                 .data = data,
 843                                 .data_end = data + hlen,
 844                         };
 845                         __be16 n_proto = proto;
 846
 847                         if (skb) {
 848                                 ctx.skb = skb;
 849                                 /* we can't use 'proto' in the skb case
 850                                  * because it might be set to skb->vlan_proto
 851                                  * which has been pulled from the data
 852                                  */
 853                                 n_proto = skb->protocol;
 854                         }
 855
 856                         ret = bpf_flow_dissect(attached, &ctx, n_proto, nhoff,
 857                                                hlen);
 858                         __skb_flow_bpf_to_target(&flow_keys, flow_dissector,
 859                                                  target_container);
 860                         rcu_read_unlock();
 861                         return ret;
 862                 }
 863                 rcu_read_unlock();
 864         }
 865
 866         if (dissector_uses_key(flow_dissector,
 867                                FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
 868                 struct ethhdr *eth = eth_hdr(skb);
 869                 struct flow_dissector_key_eth_addrs *key_eth_addrs;
 870
 871                 key_eth_addrs = skb_flow_dissector_target(flow_dissector,
 872                                                           FLOW_DISSECTOR_KEY_ETH_ADDRS,
 873                                                           target_container);
 874                 memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
 875         }
 876
 877 proto_again:
 878         fdret = FLOW_DISSECT_RET_CONTINUE;
 879
 880         switch (proto) {
 881         case htons(ETH_P_IP): {
 882                 const struct iphdr *iph;
 883                 struct iphdr _iph;
 884
 885                 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
 886                 if (!iph || iph->ihl < 5) {
 887                         fdret = FLOW_DISSECT_RET_OUT_BAD;
 888                         break;
 889                 }
 890
 891                 nhoff += iph->ihl * 4;
 892
 893                 ip_proto = iph->protocol;
 894
 895                 if (dissector_uses_key(flow_dissector,
 896                                        FLOW_DISSECTOR_KEY_IPV4_ADDRS)) {
 897                         key_addrs = skb_flow_dissector_target(flow_dissector,
 898                                                               FLOW_DISSECTOR_KEY_IPV4_ADDRS,
 899                                                               target_container);
 900
 901                         memcpy(&key_addrs->v4addrs, &iph->saddr,
 902                                sizeof(key_addrs->v4addrs));
 903                         key_control->addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
 904                 }
 905
 906                 if (ip_is_fragment(iph)) {
 907                         key_control->flags |= FLOW_DIS_IS_FRAGMENT;
 908
 909                         if (iph->frag_off & htons(IP_OFFSET)) {
 910                                 fdret = FLOW_DISSECT_RET_OUT_GOOD;
 911                                 break;
 912                         } else {
 913                                 key_control->flags |= FLOW_DIS_FIRST_FRAG;
 914                                 if (!(flags &
 915                                       FLOW_DISSECTOR_F_PARSE_1ST_FRAG)) {
 916                                         fdret = FLOW_DISSECT_RET_OUT_GOOD;
 917                                         break;
 918                                 }
 919                         }
 920                 }
 921
 922                 __skb_flow_dissect_ipv4(skb, flow_dissector,
 923                                         target_container, data, iph);
 924
 925                 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3) {
 926                         fdret = FLOW_DISSECT_RET_OUT_GOOD;
 927                         break;
 928                 }
 929
 930                 break;
 931         }
 932         case htons(ETH_P_IPV6): {
 933                 const struct ipv6hdr *iph;
 934                 struct ipv6hdr _iph;
 935
 936                 iph = __skb_header_pointer(skb, nhoff, sizeof(_iph), data, hlen, &_iph);
 937                 if (!iph) {
 938                         fdret = FLOW_DISSECT_RET_OUT_BAD;
 939                         break;
 940                 }
 941
 942                 ip_proto = iph->nexthdr;
 943                 nhoff += sizeof(struct ipv6hdr);
 944
 945                 if (dissector_uses_key(flow_dissector,
 946                                        FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
 947                         key_addrs = skb_flow_dissector_target(flow_dissector,
 948                                                               FLOW_DISSECTOR_KEY_IPV6_ADDRS,
 949                                                               target_container);
 950
 951                         memcpy(&key_addrs->v6addrs, &iph->saddr,
 952                                sizeof(key_addrs->v6addrs));
 953                         key_control->addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
 954                 }
 955
 956                 if ((dissector_uses_key(flow_dissector,
 957                                         FLOW_DISSECTOR_KEY_FLOW_LABEL) ||
 958                      (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL)) &&
 959                     ip6_flowlabel(iph)) {
 960                         __be32 flow_label = ip6_flowlabel(iph);
 961
 962                         if (dissector_uses_key(flow_dissector,
 963                                                FLOW_DISSECTOR_KEY_FLOW_LABEL)) {
 964                                 key_tags = skb_flow_dissector_target(flow_dissector,
 965                                                                      FLOW_DISSECTOR_KEY_FLOW_LABEL,
 966                                                                      target_container);
 967                                 key_tags->flow_label = ntohl(flow_label);
 968                         }
 969                         if (flags & FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL) {
 970                                 fdret = FLOW_DISSECT_RET_OUT_GOOD;
 971                                 break;
 972                         }
 973                 }
 974
 975                 __skb_flow_dissect_ipv6(skb, flow_dissector,
 976                                         target_container, data, iph);
 977
 978                 if (flags & FLOW_DISSECTOR_F_STOP_AT_L3)
 979                         fdret = FLOW_DISSECT_RET_OUT_GOOD;
 980
 981                 break;
 982         }
 983         case htons(ETH_P_8021AD):
 984         case htons(ETH_P_8021Q): {
 985                 const struct vlan_hdr *vlan = NULL;
 986                 struct vlan_hdr _vlan;
 987                 __be16 saved_vlan_tpid = proto;
 988
 989                 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX &&
 990                     skb && skb_vlan_tag_present(skb)) {
 991                         proto = skb->protocol;
 992                 } else {
 993                         vlan = __skb_header_pointer(skb, nhoff, sizeof(_vlan),
 994                                                     data, hlen, &_vlan);
 995                         if (!vlan) {
 996                                 fdret = FLOW_DISSECT_RET_OUT_BAD;
 997                                 break;
 998                         }
 999
1000                         proto = vlan->h_vlan_encapsulated_proto;
1001                         nhoff += sizeof(*vlan);
1002                 }
1003
1004                 if (dissector_vlan == FLOW_DISSECTOR_KEY_MAX) {
1005                         dissector_vlan = FLOW_DISSECTOR_KEY_VLAN;
1006                 } else if (dissector_vlan == FLOW_DISSECTOR_KEY_VLAN) {
1007                         dissector_vlan = FLOW_DISSECTOR_KEY_CVLAN;
1008                 } else {
1009                         fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1010                         break;
1011                 }
1012
1013                 if (dissector_uses_key(flow_dissector, dissector_vlan)) {
1014                         key_vlan = skb_flow_dissector_target(flow_dissector,
1015                                                              dissector_vlan,
1016                                                              target_container);
1017
1018                         if (!vlan) {
1019                                 key_vlan->vlan_id = skb_vlan_tag_get_id(skb);
1020                                 key_vlan->vlan_priority = skb_vlan_tag_get_prio(skb);
1021                         } else {
1022                                 key_vlan->vlan_id = ntohs(vlan->h_vlan_TCI) &
1023                                         VLAN_VID_MASK;
1024                                 key_vlan->vlan_priority =
1025                                         (ntohs(vlan->h_vlan_TCI) &
1026                                          VLAN_PRIO_MASK) >> VLAN_PRIO_SHIFT;
1027                         }
1028                         key_vlan->vlan_tpid = saved_vlan_tpid;
1029                 }
1030
1031                 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1032                 break;
1033         }
1034         case htons(ETH_P_PPP_SES): {
1035                 struct {
1036                         struct pppoe_hdr hdr;
1037                         __be16 proto;
1038                 } *hdr, _hdr;
1039                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
1040                 if (!hdr) {
1041                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1042                         break;
1043                 }
1044
1045                 proto = hdr->proto;
1046                 nhoff += PPPOE_SES_HLEN;
1047                 switch (proto) {
1048                 case htons(PPP_IP):
1049                         proto = htons(ETH_P_IP);
1050                         fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1051                         break;
1052                 case htons(PPP_IPV6):
1053                         proto = htons(ETH_P_IPV6);
1054                         fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1055                         break;
1056                 default:
1057                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1058                         break;
1059                 }
1060                 break;
1061         }
1062         case htons(ETH_P_TIPC): {
1063                 struct tipc_basic_hdr *hdr, _hdr;
1064
1065                 hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr),
1066                                            data, hlen, &_hdr);
1067                 if (!hdr) {
1068                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1069                         break;
1070                 }
1071
1072                 if (dissector_uses_key(flow_dissector,
1073                                        FLOW_DISSECTOR_KEY_TIPC)) {
1074                         key_addrs = skb_flow_dissector_target(flow_dissector,
1075                                                               FLOW_DISSECTOR_KEY_TIPC,
1076                                                               target_container);
1077                         key_addrs->tipckey.key = tipc_hdr_rps_key(hdr);
1078                         key_control->addr_type = FLOW_DISSECTOR_KEY_TIPC;
1079                 }
1080                 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1081                 break;
1082         }
1083
1084         case htons(ETH_P_MPLS_UC):
1085         case htons(ETH_P_MPLS_MC):
1086                 fdret = __skb_flow_dissect_mpls(skb, flow_dissector,
1087                                                 target_container, data,
1088                                                 nhoff, hlen);
1089                 break;
1090         case htons(ETH_P_FCOE):
1091                 if ((hlen - nhoff) < FCOE_HEADER_LEN) {
1092                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1093                         break;
1094                 }
1095
1096                 nhoff += FCOE_HEADER_LEN;
1097                 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1098                 break;
1099
1100         case htons(ETH_P_ARP):
1101         case htons(ETH_P_RARP):
1102                 fdret = __skb_flow_dissect_arp(skb, flow_dissector,
1103                                                target_container, data,
1104                                                nhoff, hlen);
1105                 break;
1106
1107         case htons(ETH_P_BATMAN):
1108                 fdret = __skb_flow_dissect_batadv(skb, key_control, data,
1109                                                   &proto, &nhoff, hlen, flags);
1110                 break;
1111
1112         default:
1113                 fdret = FLOW_DISSECT_RET_OUT_BAD;
1114                 break;
1115         }
1116
1117         /* Process result of proto processing */
1118         switch (fdret) {
1119         case FLOW_DISSECT_RET_OUT_GOOD:
1120                 goto out_good;
1121         case FLOW_DISSECT_RET_PROTO_AGAIN:
1122                 if (skb_flow_dissect_allowed(&num_hdrs))
1123                         goto proto_again;
1124                 goto out_good;
1125         case FLOW_DISSECT_RET_CONTINUE:
1126         case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1127                 break;
1128         case FLOW_DISSECT_RET_OUT_BAD:
1129         default:
1130                 goto out_bad;
1131         }
1132
1133 ip_proto_again:
1134         fdret = FLOW_DISSECT_RET_CONTINUE;
1135
1136         switch (ip_proto) {
1137         case IPPROTO_GRE:
1138                 fdret = __skb_flow_dissect_gre(skb, key_control, flow_dissector,
1139                                                target_container, data,
1140                                                &proto, &nhoff, &hlen, flags);
1141                 break;
1142
1143         case NEXTHDR_HOP:
1144         case NEXTHDR_ROUTING:
1145         case NEXTHDR_DEST: {
1146                 u8 _opthdr[2], *opthdr;
1147
1148                 if (proto != htons(ETH_P_IPV6))
1149                         break;
1150
1151                 opthdr = __skb_header_pointer(skb, nhoff, sizeof(_opthdr),
1152                                               data, hlen, &_opthdr);
1153                 if (!opthdr) {
1154                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1155                         break;
1156                 }
1157
1158                 ip_proto = opthdr[0];
1159                 nhoff += (opthdr[1] + 1) << 3;
1160
1161                 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1162                 break;
1163         }
1164         case NEXTHDR_FRAGMENT: {
1165                 struct frag_hdr _fh, *fh;
1166
1167                 if (proto != htons(ETH_P_IPV6))
1168                         break;
1169
1170                 fh = __skb_header_pointer(skb, nhoff, sizeof(_fh),
1171                                           data, hlen, &_fh);
1172
1173                 if (!fh) {
1174                         fdret = FLOW_DISSECT_RET_OUT_BAD;
1175                         break;
1176                 }
1177
1178                 key_control->flags |= FLOW_DIS_IS_FRAGMENT;
1179
1180                 nhoff += sizeof(_fh);
1181                 ip_proto = fh->nexthdr;
1182
1183                 if (!(fh->frag_off & htons(IP6_OFFSET))) {
1184                         key_control->flags |= FLOW_DIS_FIRST_FRAG;
1185                         if (flags & FLOW_DISSECTOR_F_PARSE_1ST_FRAG) {
1186                                 fdret = FLOW_DISSECT_RET_IPPROTO_AGAIN;
1187                                 break;
1188                         }
1189                 }
1190
1191                 fdret = FLOW_DISSECT_RET_OUT_GOOD;
1192                 break;
1193         }
1194         case IPPROTO_IPIP:
1195                 proto = htons(ETH_P_IP);
1196
1197                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1198                 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1199                         fdret = FLOW_DISSECT_RET_OUT_GOOD;
1200                         break;
1201                 }
1202
1203                 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1204                 break;
1205
1206         case IPPROTO_IPV6:
1207                 proto = htons(ETH_P_IPV6);
1208
1209                 key_control->flags |= FLOW_DIS_ENCAPSULATION;
1210                 if (flags & FLOW_DISSECTOR_F_STOP_AT_ENCAP) {
1211                         fdret = FLOW_DISSECT_RET_OUT_GOOD;
1212                         break;
1213                 }
1214
1215                 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1216                 break;
1217
1218
1219         case IPPROTO_MPLS:
1220                 proto = htons(ETH_P_MPLS_UC);
1221                 fdret = FLOW_DISSECT_RET_PROTO_AGAIN;
1222                 break;
1223
1224         case IPPROTO_TCP:
1225                 __skb_flow_dissect_tcp(skb, flow_dissector, target_container,
1226                                        data, nhoff, hlen);
1227                 break;
1228
1229         default:
1230                 break;
1231         }
1232
1233         if (dissector_uses_key(flow_dissector, FLOW_DISSECTOR_KEY_PORTS) &&
1234             !(key_control->flags & FLOW_DIS_IS_FRAGMENT)) {
1235                 key_ports = skb_flow_dissector_target(flow_dissector,
1236                                                       FLOW_DISSECTOR_KEY_PORTS,
1237                                                       target_container);
1238                 key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
1239                                                         data, hlen);
1240         }
1241
1242         if (dissector_uses_key(flow_dissector,
1243                                FLOW_DISSECTOR_KEY_ICMP)) {
1244                 key_icmp = skb_flow_dissector_target(flow_dissector,
1245                                                      FLOW_DISSECTOR_KEY_ICMP,
1246                                                      target_container);
1247                 key_icmp->icmp = skb_flow_get_be16(skb, nhoff, data, hlen);
1248         }
1249
1250         /* Process result of IP proto processing */
1251         switch (fdret) {
1252         case FLOW_DISSECT_RET_PROTO_AGAIN:
1253                 if (skb_flow_dissect_allowed(&num_hdrs))
1254                         goto proto_again;
1255                 break;
1256         case FLOW_DISSECT_RET_IPPROTO_AGAIN:
1257                 if (skb_flow_dissect_allowed(&num_hdrs))
1258                         goto ip_proto_again;
1259                 break;
1260         case FLOW_DISSECT_RET_OUT_GOOD:
1261         case FLOW_DISSECT_RET_CONTINUE:
1262                 break;
1263         case FLOW_DISSECT_RET_OUT_BAD:
1264         default:
1265                 goto out_bad;
1266         }
1267
1268 out_good:
1269         ret = true;
1270
1271 out:
1272         key_control->thoff = min_t(u16, nhoff, skb ? skb->len : hlen);
1273         key_basic->n_proto = proto;
1274         key_basic->ip_proto = ip_proto;
1275
1276         return ret;
1277
1278 out_bad:
1279         ret = false;
1280         goto out;
1281 }
1282 EXPORT_SYMBOL(__skb_flow_dissect);
1283
1284 static u32 hashrnd __read_mostly;
1285 static __always_inline void __flow_hash_secret_init(void)
1286 {
1287         net_get_random_once(&hashrnd, sizeof(hashrnd));
1288 }
1289
1290 static __always_inline u32 __flow_hash_words(const u32 *words, u32 length,
1291                                              u32 keyval)
1292 {
1293         return jhash2(words, length, keyval);
1294 }
1295
1296 static inline const u32 *flow_keys_hash_start(const struct flow_keys *flow)
1297 {
1298         const void *p = flow;
1299
1300         BUILD_BUG_ON(FLOW_KEYS_HASH_OFFSET % sizeof(u32));
1301         return (const u32 *)(p + FLOW_KEYS_HASH_OFFSET);
1302 }
1303
1304 static inline size_t flow_keys_hash_length(const struct flow_keys *flow)
1305 {
1306         size_t diff = FLOW_KEYS_HASH_OFFSET + sizeof(flow->addrs);
1307         BUILD_BUG_ON((sizeof(*flow) - FLOW_KEYS_HASH_OFFSET) % sizeof(u32));
1308         BUILD_BUG_ON(offsetof(typeof(*flow), addrs) !=
1309                      sizeof(*flow) - sizeof(flow->addrs));
1310
1311         switch (flow->control.addr_type) {
1312         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1313                 diff -= sizeof(flow->addrs.v4addrs);
1314                 break;
1315         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1316                 diff -= sizeof(flow->addrs.v6addrs);
1317                 break;
1318         case FLOW_DISSECTOR_KEY_TIPC:
1319                 diff -= sizeof(flow->addrs.tipckey);
1320                 break;
1321         }
1322         return (sizeof(*flow) - diff) / sizeof(u32);
1323 }
1324
1325 __be32 flow_get_u32_src(const struct flow_keys *flow)
1326 {
1327         switch (flow->control.addr_type) {
1328         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1329                 return flow->addrs.v4addrs.src;
1330         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1331                 return (__force __be32)ipv6_addr_hash(
1332                         &flow->addrs.v6addrs.src);
1333         case FLOW_DISSECTOR_KEY_TIPC:
1334                 return flow->addrs.tipckey.key;
1335         default:
1336                 return 0;
1337         }
1338 }
1339 EXPORT_SYMBOL(flow_get_u32_src);
1340
1341 __be32 flow_get_u32_dst(const struct flow_keys *flow)
1342 {
1343         switch (flow->control.addr_type) {
1344         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1345                 return flow->addrs.v4addrs.dst;
1346         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1347                 return (__force __be32)ipv6_addr_hash(
1348                         &flow->addrs.v6addrs.dst);
1349         default:
1350                 return 0;
1351         }
1352 }
1353 EXPORT_SYMBOL(flow_get_u32_dst);
1354
1355 static inline void __flow_hash_consistentify(struct flow_keys *keys)
1356 {
1357         int addr_diff, i;
1358
1359         switch (keys->control.addr_type) {
1360         case FLOW_DISSECTOR_KEY_IPV4_ADDRS:
1361                 addr_diff = (__force u32)keys->addrs.v4addrs.dst -
1362                             (__force u32)keys->addrs.v4addrs.src;
1363                 if ((addr_diff < 0) ||
1364                     (addr_diff == 0 &&
1365                      ((__force u16)keys->ports.dst <
1366                       (__force u16)keys->ports.src))) {
1367                         swap(keys->addrs.v4addrs.src, keys->addrs.v4addrs.dst);
1368                         swap(keys->ports.src, keys->ports.dst);
1369                 }
1370                 break;
1371         case FLOW_DISSECTOR_KEY_IPV6_ADDRS:
1372                 addr_diff = memcmp(&keys->addrs.v6addrs.dst,
1373                                    &keys->addrs.v6addrs.src,
1374                                    sizeof(keys->addrs.v6addrs.dst));
1375                 if ((addr_diff < 0) ||
1376                     (addr_diff == 0 &&
1377                      ((__force u16)keys->ports.dst <
1378                       (__force u16)keys->ports.src))) {
1379                         for (i = 0; i < 4; i++)
1380                                 swap(keys->addrs.v6addrs.src.s6_addr32[i],
1381                                      keys->addrs.v6addrs.dst.s6_addr32[i]);
1382                         swap(keys->ports.src, keys->ports.dst);
1383                 }
1384                 break;
1385         }
1386 }
1387
1388 static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
1389 {
1390         u32 hash;
1391
1392         __flow_hash_consistentify(keys);
1393
1394         hash = __flow_hash_words(flow_keys_hash_start(keys),
1395                                  flow_keys_hash_length(keys), keyval);
1396         if (!hash)
1397                 hash = 1;
1398
1399         return hash;
1400 }
1401
1402 u32 flow_hash_from_keys(struct flow_keys *keys)
1403 {
1404         __flow_hash_secret_init();
1405         return __flow_hash_from_keys(keys, hashrnd);
1406 }
1407 EXPORT_SYMBOL(flow_hash_from_keys);
1408
1409 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
1410                                   struct flow_keys *keys, u32 keyval)
1411 {
1412         skb_flow_dissect_flow_keys(skb, keys,
1413                                    FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1414
1415         return __flow_hash_from_keys(keys, keyval);
1416 }
1417
1418 struct _flow_keys_digest_data {
1419         __be16  n_proto;
1420         u8      ip_proto;
1421         u8      padding;
1422         __be32  ports;
1423         __be32  src;
1424         __be32  dst;
1425 };
1426
1427 void make_flow_keys_digest(struct flow_keys_digest *digest,
1428                            const struct flow_keys *flow)
1429 {
1430         struct _flow_keys_digest_data *data =
1431             (struct _flow_keys_digest_data *)digest;
1432
1433         BUILD_BUG_ON(sizeof(*data) > sizeof(*digest));
1434
1435         memset(digest, 0, sizeof(*digest));
1436
1437         data->n_proto = flow->basic.n_proto;
1438         data->ip_proto = flow->basic.ip_proto;
1439         data->ports = flow->ports.ports;
1440         data->src = flow->addrs.v4addrs.src;
1441         data->dst = flow->addrs.v4addrs.dst;
1442 }
1443 EXPORT_SYMBOL(make_flow_keys_digest);
1444
1445 static struct flow_dissector flow_keys_dissector_symmetric __read_mostly;
1446
1447 u32 __skb_get_hash_symmetric(const struct sk_buff *skb)
1448 {
1449         struct flow_keys keys;
1450
1451         __flow_hash_secret_init();
1452
1453         memset(&keys, 0, sizeof(keys));
1454         __skb_flow_dissect(NULL, skb, &flow_keys_dissector_symmetric,
1455                            &keys, NULL, 0, 0, 0,
1456                            FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL);
1457
1458         return __flow_hash_from_keys(&keys, hashrnd);
1459 }
1460 EXPORT_SYMBOL_GPL(__skb_get_hash_symmetric);
1461
1462 /**
1463  * __skb_get_hash: calculate a flow hash
1464  * @skb: sk_buff to calculate flow hash from
1465  *
1466  * This function calculates a flow hash based on src/dst addresses
1467  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
1468  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
1469  * if hash is a canonical 4-tuple hash over transport ports.
1470  */
1471 void __skb_get_hash(struct sk_buff *skb)
1472 {
1473         struct flow_keys keys;
1474         u32 hash;
1475
1476         __flow_hash_secret_init();
1477
1478         hash = ___skb_get_hash(skb, &keys, hashrnd);
1479
1480         __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
1481 }
1482 EXPORT_SYMBOL(__skb_get_hash);
1483
1484 __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
1485 {
1486         struct flow_keys keys;
1487
1488         return ___skb_get_hash(skb, &keys, perturb);
1489 }
1490 EXPORT_SYMBOL(skb_get_hash_perturb);
1491
1492 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1493                    const struct flow_keys_basic *keys, int hlen)
1494 {
1495         u32 poff = keys->control.thoff;
1496
1497         /* skip L4 headers for fragments after the first */
1498         if ((keys->control.flags & FLOW_DIS_IS_FRAGMENT) &&
1499             !(keys->control.flags & FLOW_DIS_FIRST_FRAG))
1500                 return poff;
1501
1502         switch (keys->basic.ip_proto) {
1503         case IPPROTO_TCP: {
1504                 /* access doff as u8 to avoid unaligned access */
1505                 const u8 *doff;
1506                 u8 _doff;
1507
1508                 doff = __skb_header_pointer(skb, poff + 12, sizeof(_doff),
1509                                             data, hlen, &_doff);
1510                 if (!doff)
1511                         return poff;
1512
1513                 poff += max_t(u32, sizeof(struct tcphdr), (*doff & 0xF0) >> 2);
1514                 break;
1515         }
1516         case IPPROTO_UDP:
1517         case IPPROTO_UDPLITE:
1518                 poff += sizeof(struct udphdr);
1519                 break;
1520         /* For the rest, we do not really care about header
1521          * extensions at this point for now.
1522          */
1523         case IPPROTO_ICMP:
1524                 poff += sizeof(struct icmphdr);
1525                 break;
1526         case IPPROTO_ICMPV6:
1527                 poff += sizeof(struct icmp6hdr);
1528                 break;
1529         case IPPROTO_IGMP:
1530                 poff += sizeof(struct igmphdr);
1531                 break;
1532         case IPPROTO_DCCP:
1533                 poff += sizeof(struct dccp_hdr);
1534                 break;
1535         case IPPROTO_SCTP:
1536                 poff += sizeof(struct sctphdr);
1537                 break;
1538         }
1539
1540         return poff;
1541 }
1542
1543 /**
1544  * skb_get_poff - get the offset to the payload
1545  * @skb: sk_buff to get the payload offset from
1546  *
1547  * The function will get the offset to the payload as far as it could
1548  * be dissected.  The main user is currently BPF, so that we can dynamically
1549  * truncate packets without needing to push actual payload to the user
1550  * space and can analyze headers only, instead.
1551  */
1552 u32 skb_get_poff(const struct sk_buff *skb)
1553 {
1554         struct flow_keys_basic keys;
1555
1556         if (!skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
1557                                               NULL, 0, 0, 0, 0))
1558                 return 0;
1559
1560         return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
1561 }
1562
1563 __u32 __get_hash_from_flowi6(const struct flowi6 *fl6, struct flow_keys *keys)
1564 {
1565         memset(keys, 0, sizeof(*keys));
1566
1567         memcpy(&keys->addrs.v6addrs.src, &fl6->saddr,
1568             sizeof(keys->addrs.v6addrs.src));
1569         memcpy(&keys->addrs.v6addrs.dst, &fl6->daddr,
1570             sizeof(keys->addrs.v6addrs.dst));
1571         keys->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
1572         keys->ports.src = fl6->fl6_sport;
1573         keys->ports.dst = fl6->fl6_dport;
1574         keys->keyid.keyid = fl6->fl6_gre_key;
1575         keys->tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
1576         keys->basic.ip_proto = fl6->flowi6_proto;
1577
1578         return flow_hash_from_keys(keys);
1579 }
1580 EXPORT_SYMBOL(__get_hash_from_flowi6);
1581
1582 static const struct flow_dissector_key flow_keys_dissector_keys[] = {
1583         {
1584                 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1585                 .offset = offsetof(struct flow_keys, control),
1586         },
1587         {
1588                 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1589                 .offset = offsetof(struct flow_keys, basic),
1590         },
1591         {
1592                 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1593                 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1594         },
1595         {
1596                 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1597                 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1598         },
1599         {
1600                 .key_id = FLOW_DISSECTOR_KEY_TIPC,
1601                 .offset = offsetof(struct flow_keys, addrs.tipckey),
1602         },
1603         {
1604                 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1605                 .offset = offsetof(struct flow_keys, ports),
1606         },
1607         {
1608                 .key_id = FLOW_DISSECTOR_KEY_VLAN,
1609                 .offset = offsetof(struct flow_keys, vlan),
1610         },
1611         {
1612                 .key_id = FLOW_DISSECTOR_KEY_FLOW_LABEL,
1613                 .offset = offsetof(struct flow_keys, tags),
1614         },
1615         {
1616                 .key_id = FLOW_DISSECTOR_KEY_GRE_KEYID,
1617                 .offset = offsetof(struct flow_keys, keyid),
1618         },
1619 };
1620
1621 static const struct flow_dissector_key flow_keys_dissector_symmetric_keys[] = {
1622         {
1623                 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1624                 .offset = offsetof(struct flow_keys, control),
1625         },
1626         {
1627                 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1628                 .offset = offsetof(struct flow_keys, basic),
1629         },
1630         {
1631                 .key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
1632                 .offset = offsetof(struct flow_keys, addrs.v4addrs),
1633         },
1634         {
1635                 .key_id = FLOW_DISSECTOR_KEY_IPV6_ADDRS,
1636                 .offset = offsetof(struct flow_keys, addrs.v6addrs),
1637         },
1638         {
1639                 .key_id = FLOW_DISSECTOR_KEY_PORTS,
1640                 .offset = offsetof(struct flow_keys, ports),
1641         },
1642 };
1643
1644 static const struct flow_dissector_key flow_keys_basic_dissector_keys[] = {
1645         {
1646                 .key_id = FLOW_DISSECTOR_KEY_CONTROL,
1647                 .offset = offsetof(struct flow_keys, control),
1648         },
1649         {
1650                 .key_id = FLOW_DISSECTOR_KEY_BASIC,
1651                 .offset = offsetof(struct flow_keys, basic),
1652         },
1653 };
1654
1655 struct flow_dissector flow_keys_dissector __read_mostly;
1656 EXPORT_SYMBOL(flow_keys_dissector);
1657
1658 struct flow_dissector flow_keys_basic_dissector __read_mostly;
1659 EXPORT_SYMBOL(flow_keys_basic_dissector);
1660
1661 static int __init init_default_flow_dissectors(void)
1662 {
1663         skb_flow_dissector_init(&flow_keys_dissector,
1664                                 flow_keys_dissector_keys,
1665                                 ARRAY_SIZE(flow_keys_dissector_keys));
1666         skb_flow_dissector_init(&flow_keys_dissector_symmetric,
1667                                 flow_keys_dissector_symmetric_keys,
1668                                 ARRAY_SIZE(flow_keys_dissector_symmetric_keys));
1669         skb_flow_dissector_init(&flow_keys_basic_dissector,
1670                                 flow_keys_basic_dissector_keys,
1671                                 ARRAY_SIZE(flow_keys_basic_dissector_keys));
1672         return 0;
1673 }
1674
1675 core_initcall(init_default_flow_dissectors);