net/openvswitch/conntrack.c

   1 /*
   2  * Copyright (c) 2015 Nicira, Inc.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  */
  13
  14 #include <linux/module.h>
  15 #include <linux/openvswitch.h>
  16 #include <linux/tcp.h>
  17 #include <linux/udp.h>
  18 #include <linux/sctp.h>
  19 #include <net/ip.h>
  20 #include <net/netfilter/nf_conntrack_core.h>
  21 #include <net/netfilter/nf_conntrack_helper.h>
  22 #include <net/netfilter/nf_conntrack_labels.h>
  23 #include <net/netfilter/nf_conntrack_seqadj.h>
  24 #include <net/netfilter/nf_conntrack_zones.h>
  25 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  26
  27 #ifdef CONFIG_NF_NAT_NEEDED
  28 #include <linux/netfilter/nf_nat.h>
  29 #include <net/netfilter/nf_nat_core.h>
  30 #include <net/netfilter/nf_nat_l3proto.h>
  31 #endif
  32
  33 #include "datapath.h"
  34 #include "conntrack.h"
  35 #include "flow.h"
  36 #include "flow_netlink.h"
  37
  38 struct ovs_ct_len_tbl {
  39         int maxlen;
  40         int minlen;
  41 };
  42
  43 /* Metadata mark for masked write to conntrack mark */
  44 struct md_mark {
  45         u32 value;
  46         u32 mask;
  47 };
  48
  49 /* Metadata label for masked write to conntrack label. */
  50 struct md_labels {
  51         struct ovs_key_ct_labels value;
  52         struct ovs_key_ct_labels mask;
  53 };
  54
  55 enum ovs_ct_nat {
  56         OVS_CT_NAT = 1 << 0,     /* NAT for committed connections only. */
  57         OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */
  58         OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */
  59 };
  60
  61 /* Conntrack action context for execution. */
  62 struct ovs_conntrack_info {
  63         struct nf_conntrack_helper *helper;
  64         struct nf_conntrack_zone zone;
  65         struct nf_conn *ct;
  66         u8 commit : 1;
  67         u8 nat : 3;                 /* enum ovs_ct_nat */
  68         u16 family;
  69         struct md_mark mark;
  70         struct md_labels labels;
  71 #ifdef CONFIG_NF_NAT_NEEDED
  72         struct nf_nat_range range;  /* Only present for SRC NAT and DST NAT. */
  73 #endif
  74 };
  75
  76 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
  77
  78 static u16 key_to_nfproto(const struct sw_flow_key *key)
  79 {
  80         switch (ntohs(key->eth.type)) {
  81         case ETH_P_IP:
  82                 return NFPROTO_IPV4;
  83         case ETH_P_IPV6:
  84                 return NFPROTO_IPV6;
  85         default:
  86                 return NFPROTO_UNSPEC;
  87         }
  88 }
  89
  90 /* Map SKB connection state into the values used by flow definition. */
  91 static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
  92 {
  93         u8 ct_state = OVS_CS_F_TRACKED;
  94
  95         switch (ctinfo) {
  96         case IP_CT_ESTABLISHED_REPLY:
  97         case IP_CT_RELATED_REPLY:
  98                 ct_state |= OVS_CS_F_REPLY_DIR;
  99                 break;
 100         default:
 101                 break;
 102         }
 103
 104         switch (ctinfo) {
 105         case IP_CT_ESTABLISHED:
 106         case IP_CT_ESTABLISHED_REPLY:
 107                 ct_state |= OVS_CS_F_ESTABLISHED;
 108                 break;
 109         case IP_CT_RELATED:
 110         case IP_CT_RELATED_REPLY:
 111                 ct_state |= OVS_CS_F_RELATED;
 112                 break;
 113         case IP_CT_NEW:
 114                 ct_state |= OVS_CS_F_NEW;
 115                 break;
 116         default:
 117                 break;
 118         }
 119
 120         return ct_state;
 121 }
 122
 123 static u32 ovs_ct_get_mark(const struct nf_conn *ct)
 124 {
 125 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
 126         return ct ? ct->mark : 0;
 127 #else
 128         return 0;
 129 #endif
 130 }
 131
 132 static void ovs_ct_get_labels(const struct nf_conn *ct,
 133                               struct ovs_key_ct_labels *labels)
 134 {
 135         struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
 136
 137         if (cl) {
 138                 size_t len = sizeof(cl->bits);
 139
 140                 if (len > OVS_CT_LABELS_LEN)
 141                         len = OVS_CT_LABELS_LEN;
 142                 else if (len < OVS_CT_LABELS_LEN)
 143                         memset(labels, 0, OVS_CT_LABELS_LEN);
 144                 memcpy(labels, cl->bits, len);
 145         } else {
 146                 memset(labels, 0, OVS_CT_LABELS_LEN);
 147         }
 148 }
 149
 150 static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
 151                                 const struct nf_conntrack_zone *zone,
 152                                 const struct nf_conn *ct)
 153 {
 154         key->ct.state = state;
 155         key->ct.zone = zone->id;
 156         key->ct.mark = ovs_ct_get_mark(ct);
 157         ovs_ct_get_labels(ct, &key->ct.labels);
 158 }
 159
 160 /* Update 'key' based on skb->nfct.  If 'post_ct' is true, then OVS has
 161  * previously sent the packet to conntrack via the ct action.  If
 162  * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
 163  * initialized from the connection status.
 164  */
 165 static void ovs_ct_update_key(const struct sk_buff *skb,
 166                               const struct ovs_conntrack_info *info,
 167                               struct sw_flow_key *key, bool post_ct,
 168                               bool keep_nat_flags)
 169 {
 170         const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
 171         enum ip_conntrack_info ctinfo;
 172         struct nf_conn *ct;
 173         u8 state = 0;
 174
 175         ct = nf_ct_get(skb, &ctinfo);
 176         if (ct) {
 177                 state = ovs_ct_get_state(ctinfo);
 178                 /* All unconfirmed entries are NEW connections. */
 179                 if (!nf_ct_is_confirmed(ct))
 180                         state |= OVS_CS_F_NEW;
 181                 /* OVS persists the related flag for the duration of the
 182                  * connection.
 183                  */
 184                 if (ct->master)
 185                         state |= OVS_CS_F_RELATED;
 186                 if (keep_nat_flags) {
 187                         state |= key->ct.state & OVS_CS_F_NAT_MASK;
 188                 } else {
 189                         if (ct->status & IPS_SRC_NAT)
 190                                 state |= OVS_CS_F_SRC_NAT;
 191                         if (ct->status & IPS_DST_NAT)
 192                                 state |= OVS_CS_F_DST_NAT;
 193                 }
 194                 zone = nf_ct_zone(ct);
 195         } else if (post_ct) {
 196                 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
 197                 if (info)
 198                         zone = &info->zone;
 199         }
 200         __ovs_ct_update_key(key, state, zone, ct);
 201 }
 202
 203 /* This is called to initialize CT key fields possibly coming in from the local
 204  * stack.
 205  */
 206 void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
 207 {
 208         ovs_ct_update_key(skb, NULL, key, false, false);
 209 }
 210
 211 int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb)
 212 {
 213         if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state))
 214                 return -EMSGSIZE;
 215
 216         if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
 217             nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone))
 218                 return -EMSGSIZE;
 219
 220         if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
 221             nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark))
 222                 return -EMSGSIZE;
 223
 224         if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
 225             nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels),
 226                     &key->ct.labels))
 227                 return -EMSGSIZE;
 228
 229         return 0;
 230 }
 231
 232 static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key,
 233                            u32 ct_mark, u32 mask)
 234 {
 235 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
 236         enum ip_conntrack_info ctinfo;
 237         struct nf_conn *ct;
 238         u32 new_mark;
 239
 240         /* The connection could be invalid, in which case set_mark is no-op. */
 241         ct = nf_ct_get(skb, &ctinfo);
 242         if (!ct)
 243                 return 0;
 244
 245         new_mark = ct_mark | (ct->mark & ~(mask));
 246         if (ct->mark != new_mark) {
 247                 ct->mark = new_mark;
 248                 nf_conntrack_event_cache(IPCT_MARK, ct);
 249                 key->ct.mark = new_mark;
 250         }
 251
 252         return 0;
 253 #else
 254         return -ENOTSUPP;
 255 #endif
 256 }
 257
 258 static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key,
 259                              const struct ovs_key_ct_labels *labels,
 260                              const struct ovs_key_ct_labels *mask)
 261 {
 262         enum ip_conntrack_info ctinfo;
 263         struct nf_conn_labels *cl;
 264         struct nf_conn *ct;
 265         int err;
 266
 267         /* The connection could be invalid, in which case set_label is no-op.*/
 268         ct = nf_ct_get(skb, &ctinfo);
 269         if (!ct)
 270                 return 0;
 271
 272         cl = nf_ct_labels_find(ct);
 273         if (!cl) {
 274                 nf_ct_labels_ext_add(ct);
 275                 cl = nf_ct_labels_find(ct);
 276         }
 277         if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN)
 278                 return -ENOSPC;
 279
 280         err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask,
 281                                     OVS_CT_LABELS_LEN / sizeof(u32));
 282         if (err)
 283                 return err;
 284
 285         ovs_ct_get_labels(ct, &key->ct.labels);
 286         return 0;
 287 }
 288
 289 /* 'skb' should already be pulled to nh_ofs. */
 290 static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
 291 {
 292         const struct nf_conntrack_helper *helper;
 293         const struct nf_conn_help *help;
 294         enum ip_conntrack_info ctinfo;
 295         unsigned int protoff;
 296         struct nf_conn *ct;
 297         int err;
 298
 299         ct = nf_ct_get(skb, &ctinfo);
 300         if (!ct || ctinfo == IP_CT_RELATED_REPLY)
 301                 return NF_ACCEPT;
 302
 303         help = nfct_help(ct);
 304         if (!help)
 305                 return NF_ACCEPT;
 306
 307         helper = rcu_dereference(help->helper);
 308         if (!helper)
 309                 return NF_ACCEPT;
 310
 311         switch (proto) {
 312         case NFPROTO_IPV4:
 313                 protoff = ip_hdrlen(skb);
 314                 break;
 315         case NFPROTO_IPV6: {
 316                 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 317                 __be16 frag_off;
 318                 int ofs;
 319
 320                 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
 321                                        &frag_off);
 322                 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
 323                         pr_debug("proto header not found\n");
 324                         return NF_ACCEPT;
 325                 }
 326                 protoff = ofs;
 327                 break;
 328         }
 329         default:
 330                 WARN_ONCE(1, "helper invoked on non-IP family!");
 331                 return NF_DROP;
 332         }
 333
 334         err = helper->help(skb, protoff, ct, ctinfo);
 335         if (err != NF_ACCEPT)
 336                 return err;
 337
 338         /* Adjust seqs after helper.  This is needed due to some helpers (e.g.,
 339          * FTP with NAT) adusting the TCP payload size when mangling IP
 340          * addresses and/or port numbers in the text-based control connection.
 341          */
 342         if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
 343             !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
 344                 return NF_DROP;
 345         return NF_ACCEPT;
 346 }
 347
 348 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
 349  * value if 'skb' is freed.
 350  */
 351 static int handle_fragments(struct net *net, struct sw_flow_key *key,
 352                             u16 zone, struct sk_buff *skb)
 353 {
 354         struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
 355         int err;
 356
 357         if (key->eth.type == htons(ETH_P_IP)) {
 358                 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
 359
 360                 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
 361                 err = ip_defrag(net, skb, user);
 362                 if (err)
 363                         return err;
 364
 365                 ovs_cb.mru = IPCB(skb)->frag_max_size;
 366 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
 367         } else if (key->eth.type == htons(ETH_P_IPV6)) {
 368                 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
 369
 370                 skb_orphan(skb);
 371                 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
 372                 err = nf_ct_frag6_gather(net, skb, user);
 373                 if (err) {
 374                         if (err != -EINPROGRESS)
 375                                 kfree_skb(skb);
 376                         return err;
 377                 }
 378
 379                 key->ip.proto = ipv6_hdr(skb)->nexthdr;
 380                 ovs_cb.mru = IP6CB(skb)->frag_max_size;
 381 #endif
 382         } else {
 383                 kfree_skb(skb);
 384                 return -EPFNOSUPPORT;
 385         }
 386
 387         key->ip.frag = OVS_FRAG_TYPE_NONE;
 388         skb_clear_hash(skb);
 389         skb->ignore_df = 1;
 390         *OVS_CB(skb) = ovs_cb;
 391
 392         return 0;
 393 }
 394
 395 static struct nf_conntrack_expect *
 396 ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
 397                    u16 proto, const struct sk_buff *skb)
 398 {
 399         struct nf_conntrack_tuple tuple;
 400
 401         if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
 402                 return NULL;
 403         return __nf_ct_expect_find(net, zone, &tuple);
 404 }
 405
 406 /* This replicates logic from nf_conntrack_core.c that is not exported. */
 407 static enum ip_conntrack_info
 408 ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
 409 {
 410         const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
 411
 412         if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
 413                 return IP_CT_ESTABLISHED_REPLY;
 414         /* Once we've had two way comms, always ESTABLISHED. */
 415         if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
 416                 return IP_CT_ESTABLISHED;
 417         if (test_bit(IPS_EXPECTED_BIT, &ct->status))
 418                 return IP_CT_RELATED;
 419         return IP_CT_NEW;
 420 }
 421
 422 /* Find an existing connection which this packet belongs to without
 423  * re-attributing statistics or modifying the connection state.  This allows an
 424  * skb->nfct lost due to an upcall to be recovered during actions execution.
 425  *
 426  * Must be called with rcu_read_lock.
 427  *
 428  * On success, populates skb->nfct and skb->nfctinfo, and returns the
 429  * connection.  Returns NULL if there is no existing entry.
 430  */
 431 static struct nf_conn *
 432 ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
 433                      u8 l3num, struct sk_buff *skb)
 434 {
 435         struct nf_conntrack_l3proto *l3proto;
 436         struct nf_conntrack_l4proto *l4proto;
 437         struct nf_conntrack_tuple tuple;
 438         struct nf_conntrack_tuple_hash *h;
 439         struct nf_conn *ct;
 440         unsigned int dataoff;
 441         u8 protonum;
 442
 443         l3proto = __nf_ct_l3proto_find(l3num);
 444         if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
 445                                  &protonum) <= 0) {
 446                 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
 447                 return NULL;
 448         }
 449         l4proto = __nf_ct_l4proto_find(l3num, protonum);
 450         if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
 451                              protonum, net, &tuple, l3proto, l4proto)) {
 452                 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
 453                 return NULL;
 454         }
 455
 456         /* look for tuple match */
 457         h = nf_conntrack_find_get(net, zone, &tuple);
 458         if (!h)
 459                 return NULL;   /* Not found. */
 460
 461         ct = nf_ct_tuplehash_to_ctrack(h);
 462
 463         skb->nfct = &ct->ct_general;
 464         skb->nfctinfo = ovs_ct_get_info(h);
 465         return ct;
 466 }
 467
 468 /* Determine whether skb->nfct is equal to the result of conntrack lookup. */
 469 static bool skb_nfct_cached(struct net *net,
 470                             const struct sw_flow_key *key,
 471                             const struct ovs_conntrack_info *info,
 472                             struct sk_buff *skb)
 473 {
 474         enum ip_conntrack_info ctinfo;
 475         struct nf_conn *ct;
 476
 477         ct = nf_ct_get(skb, &ctinfo);
 478         /* If no ct, check if we have evidence that an existing conntrack entry
 479          * might be found for this skb.  This happens when we lose a skb->nfct
 480          * due to an upcall.  If the connection was not confirmed, it is not
 481          * cached and needs to be run through conntrack again.
 482          */
 483         if (!ct && key->ct.state & OVS_CS_F_TRACKED &&
 484             !(key->ct.state & OVS_CS_F_INVALID) &&
 485             key->ct.zone == info->zone.id)
 486                 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb);
 487         if (!ct)
 488                 return false;
 489         if (!net_eq(net, read_pnet(&ct->ct_net)))
 490                 return false;
 491         if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
 492                 return false;
 493         if (info->helper) {
 494                 struct nf_conn_help *help;
 495
 496                 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
 497                 if (help && rcu_access_pointer(help->helper) != info->helper)
 498                         return false;
 499         }
 500
 501         return true;
 502 }
 503
 504 #ifdef CONFIG_NF_NAT_NEEDED
 505 /* Modelled after nf_nat_ipv[46]_fn().
 506  * range is only used for new, uninitialized NAT state.
 507  * Returns either NF_ACCEPT or NF_DROP.
 508  */
 509 static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
 510                               enum ip_conntrack_info ctinfo,
 511                               const struct nf_nat_range *range,
 512                               enum nf_nat_manip_type maniptype)
 513 {
 514         int hooknum, nh_off, err = NF_ACCEPT;
 515
 516         nh_off = skb_network_offset(skb);
 517         skb_pull(skb, nh_off);
 518
 519         /* See HOOK2MANIP(). */
 520         if (maniptype == NF_NAT_MANIP_SRC)
 521                 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
 522         else
 523                 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
 524
 525         switch (ctinfo) {
 526         case IP_CT_RELATED:
 527         case IP_CT_RELATED_REPLY:
 528                 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
 529                     skb->protocol == htons(ETH_P_IP) &&
 530                     ip_hdr(skb)->protocol == IPPROTO_ICMP) {
 531                         if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
 532                                                            hooknum))
 533                                 err = NF_DROP;
 534                         goto push;
 535                 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
 536                            skb->protocol == htons(ETH_P_IPV6)) {
 537                         __be16 frag_off;
 538                         u8 nexthdr = ipv6_hdr(skb)->nexthdr;
 539                         int hdrlen = ipv6_skip_exthdr(skb,
 540                                                       sizeof(struct ipv6hdr),
 541                                                       &nexthdr, &frag_off);
 542
 543                         if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
 544                                 if (!nf_nat_icmpv6_reply_translation(skb, ct,
 545                                                                      ctinfo,
 546                                                                      hooknum,
 547                                                                      hdrlen))
 548                                         err = NF_DROP;
 549                                 goto push;
 550                         }
 551                 }
 552                 /* Non-ICMP, fall thru to initialize if needed. */
 553         case IP_CT_NEW:
 554                 /* Seen it before?  This can happen for loopback, retrans,
 555                  * or local packets.
 556                  */
 557                 if (!nf_nat_initialized(ct, maniptype)) {
 558                         /* Initialize according to the NAT action. */
 559                         err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
 560                                 /* Action is set up to establish a new
 561                                  * mapping.
 562                                  */
 563                                 ? nf_nat_setup_info(ct, range, maniptype)
 564                                 : nf_nat_alloc_null_binding(ct, hooknum);
 565                         if (err != NF_ACCEPT)
 566                                 goto push;
 567                 }
 568                 break;
 569
 570         case IP_CT_ESTABLISHED:
 571         case IP_CT_ESTABLISHED_REPLY:
 572                 break;
 573
 574         default:
 575                 err = NF_DROP;
 576                 goto push;
 577         }
 578
 579         err = nf_nat_packet(ct, ctinfo, hooknum, skb);
 580 push:
 581         skb_push(skb, nh_off);
 582
 583         return err;
 584 }
 585
 586 static void ovs_nat_update_key(struct sw_flow_key *key,
 587                                const struct sk_buff *skb,
 588                                enum nf_nat_manip_type maniptype)
 589 {
 590         if (maniptype == NF_NAT_MANIP_SRC) {
 591                 __be16 src;
 592
 593                 key->ct.state |= OVS_CS_F_SRC_NAT;
 594                 if (key->eth.type == htons(ETH_P_IP))
 595                         key->ipv4.addr.src = ip_hdr(skb)->saddr;
 596                 else if (key->eth.type == htons(ETH_P_IPV6))
 597                         memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
 598                                sizeof(key->ipv6.addr.src));
 599                 else
 600                         return;
 601
 602                 if (key->ip.proto == IPPROTO_UDP)
 603                         src = udp_hdr(skb)->source;
 604                 else if (key->ip.proto == IPPROTO_TCP)
 605                         src = tcp_hdr(skb)->source;
 606                 else if (key->ip.proto == IPPROTO_SCTP)
 607                         src = sctp_hdr(skb)->source;
 608                 else
 609                         return;
 610
 611                 key->tp.src = src;
 612         } else {
 613                 __be16 dst;
 614
 615                 key->ct.state |= OVS_CS_F_DST_NAT;
 616                 if (key->eth.type == htons(ETH_P_IP))
 617                         key->ipv4.addr.dst = ip_hdr(skb)->daddr;
 618                 else if (key->eth.type == htons(ETH_P_IPV6))
 619                         memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
 620                                sizeof(key->ipv6.addr.dst));
 621                 else
 622                         return;
 623
 624                 if (key->ip.proto == IPPROTO_UDP)
 625                         dst = udp_hdr(skb)->dest;
 626                 else if (key->ip.proto == IPPROTO_TCP)
 627                         dst = tcp_hdr(skb)->dest;
 628                 else if (key->ip.proto == IPPROTO_SCTP)
 629                         dst = sctp_hdr(skb)->dest;
 630                 else
 631                         return;
 632
 633                 key->tp.dst = dst;
 634         }
 635 }
 636
 637 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
 638 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
 639                       const struct ovs_conntrack_info *info,
 640                       struct sk_buff *skb, struct nf_conn *ct,
 641                       enum ip_conntrack_info ctinfo)
 642 {
 643         enum nf_nat_manip_type maniptype;
 644         int err;
 645
 646         if (nf_ct_is_untracked(ct)) {
 647                 /* A NAT action may only be performed on tracked packets. */
 648                 return NF_ACCEPT;
 649         }
 650
 651         /* Add NAT extension if not confirmed yet. */
 652         if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
 653                 return NF_ACCEPT;   /* Can't NAT. */
 654
 655         /* Determine NAT type.
 656          * Check if the NAT type can be deduced from the tracked connection.
 657          * Make sure new expected connections (IP_CT_RELATED) are NATted only
 658          * when committing.
 659          */
 660         if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
 661             ct->status & IPS_NAT_MASK &&
 662             (ctinfo != IP_CT_RELATED || info->commit)) {
 663                 /* NAT an established or related connection like before. */
 664                 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
 665                         /* This is the REPLY direction for a connection
 666                          * for which NAT was applied in the forward
 667                          * direction.  Do the reverse NAT.
 668                          */
 669                         maniptype = ct->status & IPS_SRC_NAT
 670                                 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
 671                 else
 672                         maniptype = ct->status & IPS_SRC_NAT
 673                                 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
 674         } else if (info->nat & OVS_CT_SRC_NAT) {
 675                 maniptype = NF_NAT_MANIP_SRC;
 676         } else if (info->nat & OVS_CT_DST_NAT) {
 677                 maniptype = NF_NAT_MANIP_DST;
 678         } else {
 679                 return NF_ACCEPT; /* Connection is not NATed. */
 680         }
 681         err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
 682
 683         /* Mark NAT done if successful and update the flow key. */
 684         if (err == NF_ACCEPT)
 685                 ovs_nat_update_key(key, skb, maniptype);
 686
 687         return err;
 688 }
 689 #else /* !CONFIG_NF_NAT_NEEDED */
 690 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
 691                       const struct ovs_conntrack_info *info,
 692                       struct sk_buff *skb, struct nf_conn *ct,
 693                       enum ip_conntrack_info ctinfo)
 694 {
 695         return NF_ACCEPT;
 696 }
 697 #endif
 698
 699 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
 700  * not done already.  Update key with new CT state after passing the packet
 701  * through conntrack.
 702  * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
 703  * set to NULL and 0 will be returned.
 704  */
 705 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
 706                            const struct ovs_conntrack_info *info,
 707                            struct sk_buff *skb)
 708 {
 709         /* If we are recirculating packets to match on conntrack fields and
 710          * committing with a separate conntrack action,  then we don't need to
 711          * actually run the packet through conntrack twice unless it's for a
 712          * different zone.
 713          */
 714         bool cached = skb_nfct_cached(net, key, info, skb);
 715         enum ip_conntrack_info ctinfo;
 716         struct nf_conn *ct;
 717
 718         if (!cached) {
 719                 struct nf_conn *tmpl = info->ct;
 720                 int err;
 721
 722                 /* Associate skb with specified zone. */
 723                 if (tmpl) {
 724                         if (skb->nfct)
 725                                 nf_conntrack_put(skb->nfct);
 726                         nf_conntrack_get(&tmpl->ct_general);
 727                         skb->nfct = &tmpl->ct_general;
 728                         skb->nfctinfo = IP_CT_NEW;
 729                 }
 730
 731                 /* Repeat if requested, see nf_iterate(). */
 732                 do {
 733                         err = nf_conntrack_in(net, info->family,
 734                                               NF_INET_PRE_ROUTING, skb);
 735                 } while (err == NF_REPEAT);
 736
 737                 if (err != NF_ACCEPT)
 738                         return -ENOENT;
 739
 740                 /* Clear CT state NAT flags to mark that we have not yet done
 741                  * NAT after the nf_conntrack_in() call.  We can actually clear
 742                  * the whole state, as it will be re-initialized below.
 743                  */
 744                 key->ct.state = 0;
 745
 746                 /* Update the key, but keep the NAT flags. */
 747                 ovs_ct_update_key(skb, info, key, true, true);
 748         }
 749
 750         ct = nf_ct_get(skb, &ctinfo);
 751         if (ct) {
 752                 /* Packets starting a new connection must be NATted before the
 753                  * helper, so that the helper knows about the NAT.  We enforce
 754                  * this by delaying both NAT and helper calls for unconfirmed
 755                  * connections until the committing CT action.  For later
 756                  * packets NAT and Helper may be called in either order.
 757                  *
 758                  * NAT will be done only if the CT action has NAT, and only
 759                  * once per packet (per zone), as guarded by the NAT bits in
 760                  * the key->ct.state.
 761                  */
 762                 if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) &&
 763                     (nf_ct_is_confirmed(ct) || info->commit) &&
 764                     ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
 765                         return -EINVAL;
 766                 }
 767
 768                 /* Userspace may decide to perform a ct lookup without a helper
 769                  * specified followed by a (recirculate and) commit with one.
 770                  * Therefore, for unconfirmed connections which we will commit,
 771                  * we need to attach the helper here.
 772                  */
 773                 if (!nf_ct_is_confirmed(ct) && info->commit &&
 774                     info->helper && !nfct_help(ct)) {
 775                         int err = __nf_ct_try_assign_helper(ct, info->ct,
 776                                                             GFP_ATOMIC);
 777                         if (err)
 778                                 return err;
 779                 }
 780
 781                 /* Call the helper only if:
 782                  * - nf_conntrack_in() was executed above ("!cached") for a
 783                  *   confirmed connection, or
 784                  * - When committing an unconfirmed connection.
 785                  */
 786                 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
 787                     ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
 788                         return -EINVAL;
 789                 }
 790         }
 791
 792         return 0;
 793 }
 794
 795 /* Lookup connection and read fields into key. */
 796 static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
 797                          const struct ovs_conntrack_info *info,
 798                          struct sk_buff *skb)
 799 {
 800         struct nf_conntrack_expect *exp;
 801
 802         /* If we pass an expected packet through nf_conntrack_in() the
 803          * expectation is typically removed, but the packet could still be
 804          * lost in upcall processing.  To prevent this from happening we
 805          * perform an explicit expectation lookup.  Expected connections are
 806          * always new, and will be passed through conntrack only when they are
 807          * committed, as it is OK to remove the expectation at that time.
 808          */
 809         exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
 810         if (exp) {
 811                 u8 state;
 812
 813                 /* NOTE: New connections are NATted and Helped only when
 814                  * committed, so we are not calling into NAT here.
 815                  */
 816                 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
 817                 __ovs_ct_update_key(key, state, &info->zone, exp->master);
 818         } else {
 819                 struct nf_conn *ct;
 820                 int err;
 821
 822                 err = __ovs_ct_lookup(net, key, info, skb);
 823                 if (err)
 824                         return err;
 825
 826                 ct = (struct nf_conn *)skb->nfct;
 827                 if (ct)
 828                         nf_ct_deliver_cached_events(ct);
 829         }
 830
 831         return 0;
 832 }
 833
 834 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
 835 {
 836         size_t i;
 837
 838         for (i = 0; i < sizeof(*labels); i++)
 839                 if (labels->ct_labels[i])
 840                         return true;
 841
 842         return false;
 843 }
 844
 845 /* Lookup connection and confirm if unconfirmed. */
 846 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
 847                          const struct ovs_conntrack_info *info,
 848                          struct sk_buff *skb)
 849 {
 850         int err;
 851
 852         err = __ovs_ct_lookup(net, key, info, skb);
 853         if (err)
 854                 return err;
 855
 856         /* Apply changes before confirming the connection so that the initial
 857          * conntrack NEW netlink event carries the values given in the CT
 858          * action.
 859          */
 860         if (info->mark.mask) {
 861                 err = ovs_ct_set_mark(skb, key, info->mark.value,
 862                                       info->mark.mask);
 863                 if (err)
 864                         return err;
 865         }
 866         if (labels_nonzero(&info->labels.mask)) {
 867                 err = ovs_ct_set_labels(skb, key, &info->labels.value,
 868                                         &info->labels.mask);
 869                 if (err)
 870                         return err;
 871         }
 872         /* This will take care of sending queued events even if the connection
 873          * is already confirmed.
 874          */
 875         if (nf_conntrack_confirm(skb) != NF_ACCEPT)
 876                 return -EINVAL;
 877
 878         return 0;
 879 }
 880
 881 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
 882  * value if 'skb' is freed.
 883  */
 884 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
 885                    struct sw_flow_key *key,
 886                    const struct ovs_conntrack_info *info)
 887 {
 888         int nh_ofs;
 889         int err;
 890
 891         /* The conntrack module expects to be working at L3. */
 892         nh_ofs = skb_network_offset(skb);
 893         skb_pull(skb, nh_ofs);
 894
 895         if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
 896                 err = handle_fragments(net, key, info->zone.id, skb);
 897                 if (err)
 898                         return err;
 899         }
 900
 901         if (info->commit)
 902                 err = ovs_ct_commit(net, key, info, skb);
 903         else
 904                 err = ovs_ct_lookup(net, key, info, skb);
 905
 906         skb_push(skb, nh_ofs);
 907         if (err)
 908                 kfree_skb(skb);
 909         return err;
 910 }
 911
 912 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
 913                              const struct sw_flow_key *key, bool log)
 914 {
 915         struct nf_conntrack_helper *helper;
 916         struct nf_conn_help *help;
 917
 918         helper = nf_conntrack_helper_try_module_get(name, info->family,
 919                                                     key->ip.proto);
 920         if (!helper) {
 921                 OVS_NLERR(log, "Unknown helper \"%s\"", name);
 922                 return -EINVAL;
 923         }
 924
 925         help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
 926         if (!help) {
 927                 module_put(helper->me);
 928                 return -ENOMEM;
 929         }
 930
 931         rcu_assign_pointer(help->helper, helper);
 932         info->helper = helper;
 933         return 0;
 934 }
 935
 936 #ifdef CONFIG_NF_NAT_NEEDED
 937 static int parse_nat(const struct nlattr *attr,
 938                      struct ovs_conntrack_info *info, bool log)
 939 {
 940         struct nlattr *a;
 941         int rem;
 942         bool have_ip_max = false;
 943         bool have_proto_max = false;
 944         bool ip_vers = (info->family == NFPROTO_IPV6);
 945
 946         nla_for_each_nested(a, attr, rem) {
 947                 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
 948                         [OVS_NAT_ATTR_SRC] = {0, 0},
 949                         [OVS_NAT_ATTR_DST] = {0, 0},
 950                         [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
 951                                                  sizeof(struct in6_addr)},
 952                         [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
 953                                                  sizeof(struct in6_addr)},
 954                         [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
 955                         [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
 956                         [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
 957                         [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
 958                         [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
 959                 };
 960                 int type = nla_type(a);
 961
 962                 if (type > OVS_NAT_ATTR_MAX) {
 963                         OVS_NLERR(log,
 964                                   "Unknown NAT attribute (type=%d, max=%d).\n",
 965                                   type, OVS_NAT_ATTR_MAX);
 966                         return -EINVAL;
 967                 }
 968
 969                 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
 970                         OVS_NLERR(log,
 971                                   "NAT attribute type %d has unexpected length (%d != %d).\n",
 972                                   type, nla_len(a),
 973                                   ovs_nat_attr_lens[type][ip_vers]);
 974                         return -EINVAL;
 975                 }
 976
 977                 switch (type) {
 978                 case OVS_NAT_ATTR_SRC:
 979                 case OVS_NAT_ATTR_DST:
 980                         if (info->nat) {
 981                                 OVS_NLERR(log,
 982                                           "Only one type of NAT may be specified.\n"
 983                                           );
 984                                 return -ERANGE;
 985                         }
 986                         info->nat |= OVS_CT_NAT;
 987                         info->nat |= ((type == OVS_NAT_ATTR_SRC)
 988                                         ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
 989                         break;
 990
 991                 case OVS_NAT_ATTR_IP_MIN:
 992                         nla_memcpy(&info->range.min_addr, a,
 993                                    sizeof(info->range.min_addr));
 994                         info->range.flags |= NF_NAT_RANGE_MAP_IPS;
 995                         break;
 996
 997                 case OVS_NAT_ATTR_IP_MAX:
 998                         have_ip_max = true;
 999                         nla_memcpy(&info->range.max_addr, a,
1000                                    sizeof(info->range.max_addr));
1001                         info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1002                         break;
1003
1004                 case OVS_NAT_ATTR_PROTO_MIN:
1005                         info->range.min_proto.all = htons(nla_get_u16(a));
1006                         info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1007                         break;
1008
1009                 case OVS_NAT_ATTR_PROTO_MAX:
1010                         have_proto_max = true;
1011                         info->range.max_proto.all = htons(nla_get_u16(a));
1012                         info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1013                         break;
1014
1015                 case OVS_NAT_ATTR_PERSISTENT:
1016                         info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1017                         break;
1018
1019                 case OVS_NAT_ATTR_PROTO_HASH:
1020                         info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1021                         break;
1022
1023                 case OVS_NAT_ATTR_PROTO_RANDOM:
1024                         info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1025                         break;
1026
1027                 default:
1028                         OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1029                         return -EINVAL;
1030                 }
1031         }
1032
1033         if (rem > 0) {
1034                 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1035                 return -EINVAL;
1036         }
1037         if (!info->nat) {
1038                 /* Do not allow flags if no type is given. */
1039                 if (info->range.flags) {
1040                         OVS_NLERR(log,
1041                                   "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1042                                   );
1043                         return -EINVAL;
1044                 }
1045                 info->nat = OVS_CT_NAT;   /* NAT existing connections. */
1046         } else if (!info->commit) {
1047                 OVS_NLERR(log,
1048                           "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1049                           );
1050                 return -EINVAL;
1051         }
1052         /* Allow missing IP_MAX. */
1053         if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1054                 memcpy(&info->range.max_addr, &info->range.min_addr,
1055                        sizeof(info->range.max_addr));
1056         }
1057         /* Allow missing PROTO_MAX. */
1058         if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1059             !have_proto_max) {
1060                 info->range.max_proto.all = info->range.min_proto.all;
1061         }
1062         return 0;
1063 }
1064 #endif
1065
1066 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1067         [OVS_CT_ATTR_COMMIT]    = { .minlen = 0, .maxlen = 0 },
1068         [OVS_CT_ATTR_ZONE]      = { .minlen = sizeof(u16),
1069                                     .maxlen = sizeof(u16) },
1070         [OVS_CT_ATTR_MARK]      = { .minlen = sizeof(struct md_mark),
1071                                     .maxlen = sizeof(struct md_mark) },
1072         [OVS_CT_ATTR_LABELS]    = { .minlen = sizeof(struct md_labels),
1073                                     .maxlen = sizeof(struct md_labels) },
1074         [OVS_CT_ATTR_HELPER]    = { .minlen = 1,
1075                                     .maxlen = NF_CT_HELPER_NAME_LEN },
1076 #ifdef CONFIG_NF_NAT_NEEDED
1077         /* NAT length is checked when parsing the nested attributes. */
1078         [OVS_CT_ATTR_NAT]       = { .minlen = 0, .maxlen = INT_MAX },
1079 #endif
1080 };
1081
1082 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1083                     const char **helper, bool log)
1084 {
1085         struct nlattr *a;
1086         int rem;
1087
1088         nla_for_each_nested(a, attr, rem) {
1089                 int type = nla_type(a);
1090                 int maxlen = ovs_ct_attr_lens[type].maxlen;
1091                 int minlen = ovs_ct_attr_lens[type].minlen;
1092
1093                 if (type > OVS_CT_ATTR_MAX) {
1094                         OVS_NLERR(log,
1095                                   "Unknown conntrack attr (type=%d, max=%d)",
1096                                   type, OVS_CT_ATTR_MAX);
1097                         return -EINVAL;
1098                 }
1099                 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1100                         OVS_NLERR(log,
1101                                   "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1102                                   type, nla_len(a), maxlen);
1103                         return -EINVAL;
1104                 }
1105
1106                 switch (type) {
1107                 case OVS_CT_ATTR_COMMIT:
1108                         info->commit = true;
1109                         break;
1110 #ifdef CONFIG_NF_CONNTRACK_ZONES
1111                 case OVS_CT_ATTR_ZONE:
1112                         info->zone.id = nla_get_u16(a);
1113                         break;
1114 #endif
1115 #ifdef CONFIG_NF_CONNTRACK_MARK
1116                 case OVS_CT_ATTR_MARK: {
1117                         struct md_mark *mark = nla_data(a);
1118
1119                         if (!mark->mask) {
1120                                 OVS_NLERR(log, "ct_mark mask cannot be 0");
1121                                 return -EINVAL;
1122                         }
1123                         info->mark = *mark;
1124                         break;
1125                 }
1126 #endif
1127 #ifdef CONFIG_NF_CONNTRACK_LABELS
1128                 case OVS_CT_ATTR_LABELS: {
1129                         struct md_labels *labels = nla_data(a);
1130
1131                         if (!labels_nonzero(&labels->mask)) {
1132                                 OVS_NLERR(log, "ct_labels mask cannot be 0");
1133                                 return -EINVAL;
1134                         }
1135                         info->labels = *labels;
1136                         break;
1137                 }
1138 #endif
1139                 case OVS_CT_ATTR_HELPER:
1140                         *helper = nla_data(a);
1141                         if (!memchr(*helper, '\0', nla_len(a))) {
1142                                 OVS_NLERR(log, "Invalid conntrack helper");
1143                                 return -EINVAL;
1144                         }
1145                         break;
1146 #ifdef CONFIG_NF_NAT_NEEDED
1147                 case OVS_CT_ATTR_NAT: {
1148                         int err = parse_nat(a, info, log);
1149
1150                         if (err)
1151                                 return err;
1152                         break;
1153                 }
1154 #endif
1155                 default:
1156                         OVS_NLERR(log, "Unknown conntrack attr (%d)",
1157                                   type);
1158                         return -EINVAL;
1159                 }
1160         }
1161
1162 #ifdef CONFIG_NF_CONNTRACK_MARK
1163         if (!info->commit && info->mark.mask) {
1164                 OVS_NLERR(log,
1165                           "Setting conntrack mark requires 'commit' flag.");
1166                 return -EINVAL;
1167         }
1168 #endif
1169 #ifdef CONFIG_NF_CONNTRACK_LABELS
1170         if (!info->commit && labels_nonzero(&info->labels.mask)) {
1171                 OVS_NLERR(log,
1172                           "Setting conntrack labels requires 'commit' flag.");
1173                 return -EINVAL;
1174         }
1175 #endif
1176         if (rem > 0) {
1177                 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1178                 return -EINVAL;
1179         }
1180
1181         return 0;
1182 }
1183
1184 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1185 {
1186         if (attr == OVS_KEY_ATTR_CT_STATE)
1187                 return true;
1188         if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1189             attr == OVS_KEY_ATTR_CT_ZONE)
1190                 return true;
1191         if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1192             attr == OVS_KEY_ATTR_CT_MARK)
1193                 return true;
1194         if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1195             attr == OVS_KEY_ATTR_CT_LABELS) {
1196                 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1197
1198                 return ovs_net->xt_label;
1199         }
1200
1201         return false;
1202 }
1203
1204 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1205                        const struct sw_flow_key *key,
1206                        struct sw_flow_actions **sfa,  bool log)
1207 {
1208         struct ovs_conntrack_info ct_info;
1209         const char *helper = NULL;
1210         u16 family;
1211         int err;
1212
1213         family = key_to_nfproto(key);
1214         if (family == NFPROTO_UNSPEC) {
1215                 OVS_NLERR(log, "ct family unspecified");
1216                 return -EINVAL;
1217         }
1218
1219         memset(&ct_info, 0, sizeof(ct_info));
1220         ct_info.family = family;
1221
1222         nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1223                         NF_CT_DEFAULT_ZONE_DIR, 0);
1224
1225         err = parse_ct(attr, &ct_info, &helper, log);
1226         if (err)
1227                 return err;
1228
1229         /* Set up template for tracking connections in specific zones. */
1230         ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1231         if (!ct_info.ct) {
1232                 OVS_NLERR(log, "Failed to allocate conntrack template");
1233                 return -ENOMEM;
1234         }
1235
1236         __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1237         nf_conntrack_get(&ct_info.ct->ct_general);
1238
1239         if (helper) {
1240                 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1241                 if (err)
1242                         goto err_free_ct;
1243         }
1244
1245         err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1246                                  sizeof(ct_info), log);
1247         if (err)
1248                 goto err_free_ct;
1249
1250         return 0;
1251 err_free_ct:
1252         __ovs_ct_free_action(&ct_info);
1253         return err;
1254 }
1255
1256 #ifdef CONFIG_NF_NAT_NEEDED
1257 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1258                                struct sk_buff *skb)
1259 {
1260         struct nlattr *start;
1261
1262         start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1263         if (!start)
1264                 return false;
1265
1266         if (info->nat & OVS_CT_SRC_NAT) {
1267                 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1268                         return false;
1269         } else if (info->nat & OVS_CT_DST_NAT) {
1270                 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1271                         return false;
1272         } else {
1273                 goto out;
1274         }
1275
1276         if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1277                 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1278                     info->family == NFPROTO_IPV4) {
1279                         if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1280                                             info->range.min_addr.ip) ||
1281                             (info->range.max_addr.ip
1282                              != info->range.min_addr.ip &&
1283                              (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1284                                               info->range.max_addr.ip))))
1285                                 return false;
1286                 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1287                            info->family == NFPROTO_IPV6) {
1288                         if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1289                                              &info->range.min_addr.in6) ||
1290                             (memcmp(&info->range.max_addr.in6,
1291                                     &info->range.min_addr.in6,
1292                                     sizeof(info->range.max_addr.in6)) &&
1293                              (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1294                                                &info->range.max_addr.in6))))
1295                                 return false;
1296                 } else {
1297                         return false;
1298                 }
1299         }
1300         if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1301             (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1302                          ntohs(info->range.min_proto.all)) ||
1303              (info->range.max_proto.all != info->range.min_proto.all &&
1304               nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1305                           ntohs(info->range.max_proto.all)))))
1306                 return false;
1307
1308         if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1309             nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1310                 return false;
1311         if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1312             nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1313                 return false;
1314         if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1315             nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1316                 return false;
1317 out:
1318         nla_nest_end(skb, start);
1319
1320         return true;
1321 }
1322 #endif
1323
1324 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1325                           struct sk_buff *skb)
1326 {
1327         struct nlattr *start;
1328
1329         start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1330         if (!start)
1331                 return -EMSGSIZE;
1332
1333         if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
1334                 return -EMSGSIZE;
1335         if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1336             nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1337                 return -EMSGSIZE;
1338         if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1339             nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1340                     &ct_info->mark))
1341                 return -EMSGSIZE;
1342         if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1343             labels_nonzero(&ct_info->labels.mask) &&
1344             nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1345                     &ct_info->labels))
1346                 return -EMSGSIZE;
1347         if (ct_info->helper) {
1348                 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1349                                    ct_info->helper->name))
1350                         return -EMSGSIZE;
1351         }
1352 #ifdef CONFIG_NF_NAT_NEEDED
1353         if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1354                 return -EMSGSIZE;
1355 #endif
1356         nla_nest_end(skb, start);
1357
1358         return 0;
1359 }
1360
1361 void ovs_ct_free_action(const struct nlattr *a)
1362 {
1363         struct ovs_conntrack_info *ct_info = nla_data(a);
1364
1365         __ovs_ct_free_action(ct_info);
1366 }
1367
1368 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1369 {
1370         if (ct_info->helper)
1371                 module_put(ct_info->helper->me);
1372         if (ct_info->ct)
1373                 nf_ct_tmpl_free(ct_info->ct);
1374 }
1375
1376 void ovs_ct_init(struct net *net)
1377 {
1378         unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
1379         struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1380
1381         if (nf_connlabels_get(net, n_bits - 1)) {
1382                 ovs_net->xt_label = false;
1383                 OVS_NLERR(true, "Failed to set connlabel length");
1384         } else {
1385                 ovs_net->xt_label = true;
1386         }
1387 }
1388
1389 void ovs_ct_exit(struct net *net)
1390 {
1391         struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1392
1393         if (ovs_net->xt_label)
1394                 nf_connlabels_put(net);
1395 }