net/ipv4/netfilter/ip_conntrack_netlink.c

   1 /* Connection tracking via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
   7  * (C) 2005-2006 by Pablo Neira Ayuso <pablo@eurodev.net>
   8  *
   9  * I've reworked this stuff to use attributes instead of conntrack
  10  * structures. 5.44 am. I need more tea. --pablo 05/07/11.
  11  *
  12  * Initial connection tracking via netlink development funded and
  13  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
  14  *
  15  * Further development of this code funded by Astaro AG (http://www.astaro.com)
  16  *
  17  * This software may be used and distributed according to the terms
  18  * of the GNU General Public License, incorporated herein by reference.
  19  */
  20
  21 #include <linux/init.h>
  22 #include <linux/module.h>
  23 #include <linux/kernel.h>
  24 #include <linux/types.h>
  25 #include <linux/timer.h>
  26 #include <linux/skbuff.h>
  27 #include <linux/errno.h>
  28 #include <linux/netlink.h>
  29 #include <linux/spinlock.h>
  30 #include <linux/interrupt.h>
  31 #include <linux/notifier.h>
  32
  33 #include <linux/netfilter.h>
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  38 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  39
  40 #include <linux/netfilter/nfnetlink.h>
  41 #include <linux/netfilter/nfnetlink_conntrack.h>
  42
  43 MODULE_LICENSE("GPL");
  44
  45 static char __initdata version[] = "0.90";
  46
  47 static inline int
  48 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  49                             const struct ip_conntrack_tuple *tuple,
  50                             struct ip_conntrack_protocol *proto)
  51 {
  52         int ret = 0;
  53         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  54
  55         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  56
  57         if (likely(proto->tuple_to_nfattr))
  58                 ret = proto->tuple_to_nfattr(skb, tuple);
  59
  60         NFA_NEST_END(skb, nest_parms);
  61
  62         return ret;
  63
  64 nfattr_failure:
  65         return -1;
  66 }
  67
  68 static inline int
  69 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
  70                          const struct ip_conntrack_tuple *tuple)
  71 {
  72         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  73
  74         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(__be32), &tuple->src.ip);
  75         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(__be32), &tuple->dst.ip);
  76
  77         NFA_NEST_END(skb, nest_parms);
  78
  79         return 0;
  80
  81 nfattr_failure:
  82         return -1;
  83 }
  84
  85 static inline int
  86 ctnetlink_dump_tuples(struct sk_buff *skb,
  87                       const struct ip_conntrack_tuple *tuple)
  88 {
  89         int ret;
  90         struct ip_conntrack_protocol *proto;
  91
  92         ret = ctnetlink_dump_tuples_ip(skb, tuple);
  93         if (unlikely(ret < 0))
  94                 return ret;
  95
  96         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  97         ret = ctnetlink_dump_tuples_proto(skb, tuple, proto);
  98         ip_conntrack_proto_put(proto);
  99
 100         return ret;
 101 }
 102
 103 static inline int
 104 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 105 {
 106         __be32 status = htonl((u_int32_t) ct->status);
 107         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 108         return 0;
 109
 110 nfattr_failure:
 111         return -1;
 112 }
 113
 114 static inline int
 115 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 116 {
 117         long timeout_l = ct->timeout.expires - jiffies;
 118         __be32 timeout;
 119
 120         if (timeout_l < 0)
 121                 timeout = 0;
 122         else
 123                 timeout = htonl(timeout_l / HZ);
 124
 125         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 126         return 0;
 127
 128 nfattr_failure:
 129         return -1;
 130 }
 131
 132 static inline int
 133 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 134 {
 135         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 136
 137         struct nfattr *nest_proto;
 138         int ret;
 139
 140         if (!proto->to_nfattr) {
 141                 ip_conntrack_proto_put(proto);
 142                 return 0;
 143         }
 144
 145         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 146
 147         ret = proto->to_nfattr(skb, nest_proto, ct);
 148
 149         ip_conntrack_proto_put(proto);
 150
 151         NFA_NEST_END(skb, nest_proto);
 152
 153         return ret;
 154
 155 nfattr_failure:
 156         ip_conntrack_proto_put(proto);
 157         return -1;
 158 }
 159
 160 static inline int
 161 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 162 {
 163         struct nfattr *nest_helper;
 164
 165         if (!ct->helper)
 166                 return 0;
 167
 168         nest_helper = NFA_NEST(skb, CTA_HELP);
 169         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 170
 171         if (ct->helper->to_nfattr)
 172                 ct->helper->to_nfattr(skb, ct);
 173
 174         NFA_NEST_END(skb, nest_helper);
 175
 176         return 0;
 177
 178 nfattr_failure:
 179         return -1;
 180 }
 181
 182 #ifdef CONFIG_IP_NF_CT_ACCT
 183 static inline int
 184 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 185                         enum ip_conntrack_dir dir)
 186 {
 187         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 188         struct nfattr *nest_count = NFA_NEST(skb, type);
 189         __be32 tmp;
 190
 191         tmp = htonl(ct->counters[dir].packets);
 192         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(__be32), &tmp);
 193
 194         tmp = htonl(ct->counters[dir].bytes);
 195         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(__be32), &tmp);
 196
 197         NFA_NEST_END(skb, nest_count);
 198
 199         return 0;
 200
 201 nfattr_failure:
 202         return -1;
 203 }
 204 #else
 205 #define ctnetlink_dump_counters(a, b, c) (0)
 206 #endif
 207
 208 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 209 static inline int
 210 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 211 {
 212         __be32 mark = htonl(ct->mark);
 213
 214         NFA_PUT(skb, CTA_MARK, sizeof(__be32), &mark);
 215         return 0;
 216
 217 nfattr_failure:
 218         return -1;
 219 }
 220 #else
 221 #define ctnetlink_dump_mark(a, b) (0)
 222 #endif
 223
 224 static inline int
 225 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 226 {
 227         __be32 id = htonl(ct->id);
 228         NFA_PUT(skb, CTA_ID, sizeof(__be32), &id);
 229         return 0;
 230
 231 nfattr_failure:
 232         return -1;
 233 }
 234
 235 static inline int
 236 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 237 {
 238         __be32 use = htonl(atomic_read(&ct->ct_general.use));
 239
 240         NFA_PUT(skb, CTA_USE, sizeof(__be32), &use);
 241         return 0;
 242
 243 nfattr_failure:
 244         return -1;
 245 }
 246
 247 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 248
 249 static int
 250 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 251                     int event, int nowait,
 252                     const struct ip_conntrack *ct)
 253 {
 254         struct nlmsghdr *nlh;
 255         struct nfgenmsg *nfmsg;
 256         struct nfattr *nest_parms;
 257         unsigned char *b;
 258
 259         b = skb->tail;
 260
 261         event |= NFNL_SUBSYS_CTNETLINK << 8;
 262         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 263         nfmsg  = NLMSG_DATA(nlh);
 264
 265         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 266         nfmsg->nfgen_family = AF_INET;
 267         nfmsg->version      = NFNETLINK_V0;
 268         nfmsg->res_id       = 0;
 269
 270         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 271         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 272                 goto nfattr_failure;
 273         NFA_NEST_END(skb, nest_parms);
 274
 275         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 276         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 277                 goto nfattr_failure;
 278         NFA_NEST_END(skb, nest_parms);
 279
 280         if (ctnetlink_dump_status(skb, ct) < 0 ||
 281             ctnetlink_dump_timeout(skb, ct) < 0 ||
 282             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 283             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 284             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 285             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 286             ctnetlink_dump_mark(skb, ct) < 0 ||
 287             ctnetlink_dump_id(skb, ct) < 0 ||
 288             ctnetlink_dump_use(skb, ct) < 0)
 289                 goto nfattr_failure;
 290
 291         nlh->nlmsg_len = skb->tail - b;
 292         return skb->len;
 293
 294 nlmsg_failure:
 295 nfattr_failure:
 296         skb_trim(skb, b - skb->data);
 297         return -1;
 298 }
 299
 300 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 301 static int ctnetlink_conntrack_event(struct notifier_block *this,
 302                                      unsigned long events, void *ptr)
 303 {
 304         struct nlmsghdr *nlh;
 305         struct nfgenmsg *nfmsg;
 306         struct nfattr *nest_parms;
 307         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 308         struct sk_buff *skb;
 309         unsigned int type;
 310         unsigned char *b;
 311         unsigned int flags = 0, group;
 312
 313         /* ignore our fake conntrack entry */
 314         if (ct == &ip_conntrack_untracked)
 315                 return NOTIFY_DONE;
 316
 317         if (events & IPCT_DESTROY) {
 318                 type = IPCTNL_MSG_CT_DELETE;
 319                 group = NFNLGRP_CONNTRACK_DESTROY;
 320         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 321                 type = IPCTNL_MSG_CT_NEW;
 322                 flags = NLM_F_CREATE|NLM_F_EXCL;
 323                 group = NFNLGRP_CONNTRACK_NEW;
 324         } else if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
 325                 type = IPCTNL_MSG_CT_NEW;
 326                 group = NFNLGRP_CONNTRACK_UPDATE;
 327         } else
 328                 return NOTIFY_DONE;
 329
 330         if (!nfnetlink_has_listeners(group))
 331                 return NOTIFY_DONE;
 332
 333         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 334         if (!skb)
 335                 return NOTIFY_DONE;
 336
 337         b = skb->tail;
 338
 339         type |= NFNL_SUBSYS_CTNETLINK << 8;
 340         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 341         nfmsg = NLMSG_DATA(nlh);
 342
 343         nlh->nlmsg_flags    = flags;
 344         nfmsg->nfgen_family = AF_INET;
 345         nfmsg->version  = NFNETLINK_V0;
 346         nfmsg->res_id   = 0;
 347
 348         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 349         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 350                 goto nfattr_failure;
 351         NFA_NEST_END(skb, nest_parms);
 352
 353         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 354         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 355                 goto nfattr_failure;
 356         NFA_NEST_END(skb, nest_parms);
 357
 358         if (events & IPCT_DESTROY) {
 359                 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 360                     ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 361                         goto nfattr_failure;
 362         } else {
 363                 if (ctnetlink_dump_status(skb, ct) < 0)
 364                         goto nfattr_failure;
 365
 366                 if (ctnetlink_dump_timeout(skb, ct) < 0)
 367                         goto nfattr_failure;
 368
 369                 if (events & IPCT_PROTOINFO
 370                     && ctnetlink_dump_protoinfo(skb, ct) < 0)
 371                         goto nfattr_failure;
 372
 373                 if ((events & IPCT_HELPER || ct->helper)
 374                     && ctnetlink_dump_helpinfo(skb, ct) < 0)
 375                         goto nfattr_failure;
 376
 377                 if ((events & IPCT_MARK || ct->mark)
 378                     && ctnetlink_dump_mark(skb, ct) < 0)
 379                         goto nfattr_failure;
 380
 381                 if (events & IPCT_COUNTER_FILLING &&
 382                     (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 383                      ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0))
 384                         goto nfattr_failure;
 385         }
 386
 387         nlh->nlmsg_len = skb->tail - b;
 388         nfnetlink_send(skb, 0, group, 0);
 389         return NOTIFY_DONE;
 390
 391 nlmsg_failure:
 392 nfattr_failure:
 393         kfree_skb(skb);
 394         return NOTIFY_DONE;
 395 }
 396 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 397
 398 static int ctnetlink_done(struct netlink_callback *cb)
 399 {
 400         if (cb->args[1])
 401                 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
 402         return 0;
 403 }
 404
 405 static int
 406 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 407 {
 408         struct ip_conntrack *ct, *last;
 409         struct ip_conntrack_tuple_hash *h;
 410         struct list_head *i;
 411
 412         read_lock_bh(&ip_conntrack_lock);
 413         last = (struct ip_conntrack *)cb->args[1];
 414         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
 415 restart:
 416                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 417                         h = (struct ip_conntrack_tuple_hash *) i;
 418                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 419                                 continue;
 420                         ct = tuplehash_to_ctrack(h);
 421                         if (cb->args[1]) {
 422                                 if (ct != last)
 423                                         continue;
 424                                 cb->args[1] = 0;
 425                         }
 426                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 427                                                 cb->nlh->nlmsg_seq,
 428                                                 IPCTNL_MSG_CT_NEW,
 429                                                 1, ct) < 0) {
 430                                 nf_conntrack_get(&ct->ct_general);
 431                                 cb->args[1] = (unsigned long)ct;
 432                                 goto out;
 433                         }
 434 #ifdef CONFIG_NF_CT_ACCT
 435                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
 436                                                 IPCTNL_MSG_CT_GET_CTRZERO)
 437                                 memset(&ct->counters, 0, sizeof(ct->counters));
 438 #endif
 439                 }
 440                 if (cb->args[1]) {
 441                         cb->args[1] = 0;
 442                         goto restart;
 443                 }
 444         }
 445 out:
 446         read_unlock_bh(&ip_conntrack_lock);
 447         if (last)
 448                 ip_conntrack_put(last);
 449
 450         return skb->len;
 451 }
 452
 453 static const size_t cta_min_ip[CTA_IP_MAX] = {
 454         [CTA_IP_V4_SRC-1]       = sizeof(__be32),
 455         [CTA_IP_V4_DST-1]       = sizeof(__be32),
 456 };
 457
 458 static inline int
 459 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 460 {
 461         struct nfattr *tb[CTA_IP_MAX];
 462
 463         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 464
 465         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 466                 return -EINVAL;
 467
 468         if (!tb[CTA_IP_V4_SRC-1])
 469                 return -EINVAL;
 470         tuple->src.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 471
 472         if (!tb[CTA_IP_V4_DST-1])
 473                 return -EINVAL;
 474         tuple->dst.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 475
 476         return 0;
 477 }
 478
 479 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 480         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 481         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 482         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 483         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 484         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 485         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 486 };
 487
 488 static inline int
 489 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 490                             struct ip_conntrack_tuple *tuple)
 491 {
 492         struct nfattr *tb[CTA_PROTO_MAX];
 493         struct ip_conntrack_protocol *proto;
 494         int ret = 0;
 495
 496         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 497
 498         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 499                 return -EINVAL;
 500
 501         if (!tb[CTA_PROTO_NUM-1])
 502                 return -EINVAL;
 503         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 504
 505         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 506
 507         if (likely(proto->nfattr_to_tuple))
 508                 ret = proto->nfattr_to_tuple(tb, tuple);
 509
 510         ip_conntrack_proto_put(proto);
 511
 512         return ret;
 513 }
 514
 515 static inline int
 516 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 517                       enum ctattr_tuple type)
 518 {
 519         struct nfattr *tb[CTA_TUPLE_MAX];
 520         int err;
 521
 522         memset(tuple, 0, sizeof(*tuple));
 523
 524         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 525
 526         if (!tb[CTA_TUPLE_IP-1])
 527                 return -EINVAL;
 528
 529         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 530         if (err < 0)
 531                 return err;
 532
 533         if (!tb[CTA_TUPLE_PROTO-1])
 534                 return -EINVAL;
 535
 536         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 537         if (err < 0)
 538                 return err;
 539
 540         /* orig and expect tuples get DIR_ORIGINAL */
 541         if (type == CTA_TUPLE_REPLY)
 542                 tuple->dst.dir = IP_CT_DIR_REPLY;
 543         else
 544                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 545
 546         return 0;
 547 }
 548
 549 #ifdef CONFIG_IP_NF_NAT_NEEDED
 550 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 551         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 552         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 553 };
 554
 555 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 556                                      const struct ip_conntrack *ct,
 557                                      struct ip_nat_range *range)
 558 {
 559         struct nfattr *tb[CTA_PROTONAT_MAX];
 560         struct ip_nat_protocol *npt;
 561
 562         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 563
 564         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 565                 return -EINVAL;
 566
 567         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 568
 569         if (!npt->nfattr_to_range) {
 570                 ip_nat_proto_put(npt);
 571                 return 0;
 572         }
 573
 574         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 575         if (npt->nfattr_to_range(tb, range) > 0)
 576                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 577
 578         ip_nat_proto_put(npt);
 579
 580         return 0;
 581 }
 582
 583 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 584         [CTA_NAT_MINIP-1]       = sizeof(__be32),
 585         [CTA_NAT_MAXIP-1]       = sizeof(__be32),
 586 };
 587
 588 static inline int
 589 ctnetlink_parse_nat(struct nfattr *nat,
 590                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 591 {
 592         struct nfattr *tb[CTA_NAT_MAX];
 593         int err;
 594
 595         memset(range, 0, sizeof(*range));
 596
 597         nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
 598
 599         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 600                 return -EINVAL;
 601
 602         if (tb[CTA_NAT_MINIP-1])
 603                 range->min_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 604
 605         if (!tb[CTA_NAT_MAXIP-1])
 606                 range->max_ip = range->min_ip;
 607         else
 608                 range->max_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 609
 610         if (range->min_ip)
 611                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 612
 613         if (!tb[CTA_NAT_PROTO-1])
 614                 return 0;
 615
 616         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 617         if (err < 0)
 618                 return err;
 619
 620         return 0;
 621 }
 622 #endif
 623
 624 static inline int
 625 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 626 {
 627         struct nfattr *tb[CTA_HELP_MAX];
 628
 629         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 630
 631         if (!tb[CTA_HELP_NAME-1])
 632                 return -EINVAL;
 633
 634         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 635
 636         return 0;
 637 }
 638
 639 static const size_t cta_min[CTA_MAX] = {
 640         [CTA_STATUS-1]          = sizeof(__be32),
 641         [CTA_TIMEOUT-1]         = sizeof(__be32),
 642         [CTA_MARK-1]            = sizeof(__be32),
 643         [CTA_USE-1]             = sizeof(__be32),
 644         [CTA_ID-1]              = sizeof(__be32)
 645 };
 646
 647 static int
 648 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 649                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 650 {
 651         struct ip_conntrack_tuple_hash *h;
 652         struct ip_conntrack_tuple tuple;
 653         struct ip_conntrack *ct;
 654         int err = 0;
 655
 656         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 657                 return -EINVAL;
 658
 659         if (cda[CTA_TUPLE_ORIG-1])
 660                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 661         else if (cda[CTA_TUPLE_REPLY-1])
 662                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 663         else {
 664                 /* Flush the whole table */
 665                 ip_conntrack_flush();
 666                 return 0;
 667         }
 668
 669         if (err < 0)
 670                 return err;
 671
 672         h = ip_conntrack_find_get(&tuple, NULL);
 673         if (!h)
 674                 return -ENOENT;
 675
 676         ct = tuplehash_to_ctrack(h);
 677
 678         if (cda[CTA_ID-1]) {
 679                 u_int32_t id = ntohl(*(__be32 *)NFA_DATA(cda[CTA_ID-1]));
 680                 if (ct->id != id) {
 681                         ip_conntrack_put(ct);
 682                         return -ENOENT;
 683                 }
 684         }
 685         if (del_timer(&ct->timeout))
 686                 ct->timeout.function((unsigned long)ct);
 687
 688         ip_conntrack_put(ct);
 689
 690         return 0;
 691 }
 692
 693 static int
 694 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 695                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 696 {
 697         struct ip_conntrack_tuple_hash *h;
 698         struct ip_conntrack_tuple tuple;
 699         struct ip_conntrack *ct;
 700         struct sk_buff *skb2 = NULL;
 701         int err = 0;
 702
 703         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 704                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 705                 u32 rlen;
 706
 707                 if (msg->nfgen_family != AF_INET)
 708                         return -EAFNOSUPPORT;
 709
 710 #ifndef CONFIG_IP_NF_CT_ACCT
 711                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
 712                         return -ENOTSUPP;
 713 #endif
 714                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 715                                                 ctnetlink_dump_table,
 716                                                 ctnetlink_done)) != 0)
 717                         return -EINVAL;
 718
 719                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 720                 if (rlen > skb->len)
 721                         rlen = skb->len;
 722                 skb_pull(skb, rlen);
 723                 return 0;
 724         }
 725
 726         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 727                 return -EINVAL;
 728
 729         if (cda[CTA_TUPLE_ORIG-1])
 730                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 731         else if (cda[CTA_TUPLE_REPLY-1])
 732                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 733         else
 734                 return -EINVAL;
 735
 736         if (err < 0)
 737                 return err;
 738
 739         h = ip_conntrack_find_get(&tuple, NULL);
 740         if (!h)
 741                 return -ENOENT;
 742
 743         ct = tuplehash_to_ctrack(h);
 744
 745         err = -ENOMEM;
 746         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 747         if (!skb2) {
 748                 ip_conntrack_put(ct);
 749                 return -ENOMEM;
 750         }
 751
 752         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 753                                   IPCTNL_MSG_CT_NEW, 1, ct);
 754         ip_conntrack_put(ct);
 755         if (err <= 0)
 756                 goto free;
 757
 758         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 759         if (err < 0)
 760                 goto out;
 761
 762         return 0;
 763
 764 free:
 765         kfree_skb(skb2);
 766 out:
 767         return err;
 768 }
 769
 770 static inline int
 771 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 772 {
 773         unsigned long d;
 774         unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
 775         d = ct->status ^ status;
 776
 777         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 778                 /* unchangeable */
 779                 return -EINVAL;
 780
 781         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 782                 /* SEEN_REPLY bit can only be set */
 783                 return -EINVAL;
 784
 785
 786         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 787                 /* ASSURED bit can only be set */
 788                 return -EINVAL;
 789
 790         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 791 #ifndef CONFIG_IP_NF_NAT_NEEDED
 792                 return -EINVAL;
 793 #else
 794                 struct ip_nat_range range;
 795
 796                 if (cda[CTA_NAT_DST-1]) {
 797                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 798                                                 &range) < 0)
 799                                 return -EINVAL;
 800                         if (ip_nat_initialized(ct,
 801                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 802                                 return -EEXIST;
 803                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 804                 }
 805                 if (cda[CTA_NAT_SRC-1]) {
 806                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 807                                                 &range) < 0)
 808                                 return -EINVAL;
 809                         if (ip_nat_initialized(ct,
 810                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 811                                 return -EEXIST;
 812                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 813                 }
 814 #endif
 815         }
 816
 817         /* Be careful here, modifying NAT bits can screw up things,
 818          * so don't let users modify them directly if they don't pass
 819          * ip_nat_range. */
 820         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 821         return 0;
 822 }
 823
 824
 825 static inline int
 826 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 827 {
 828         struct ip_conntrack_helper *helper;
 829         char *helpname;
 830         int err;
 831
 832         /* don't change helper of sibling connections */
 833         if (ct->master)
 834                 return -EINVAL;
 835
 836         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 837         if (err < 0)
 838                 return err;
 839
 840         helper = __ip_conntrack_helper_find_byname(helpname);
 841         if (!helper) {
 842                 if (!strcmp(helpname, ""))
 843                         helper = NULL;
 844                 else
 845                         return -EINVAL;
 846         }
 847
 848         if (ct->helper) {
 849                 if (!helper) {
 850                         /* we had a helper before ... */
 851                         ip_ct_remove_expectations(ct);
 852                         ct->helper = NULL;
 853                 } else {
 854                         /* need to zero data of old helper */
 855                         memset(&ct->help, 0, sizeof(ct->help));
 856                 }
 857         }
 858
 859         ct->helper = helper;
 860
 861         return 0;
 862 }
 863
 864 static inline int
 865 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 866 {
 867         u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 868
 869         if (!del_timer(&ct->timeout))
 870                 return -ETIME;
 871
 872         ct->timeout.expires = jiffies + timeout * HZ;
 873         add_timer(&ct->timeout);
 874
 875         return 0;
 876 }
 877
 878 static inline int
 879 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 880 {
 881         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 882         struct ip_conntrack_protocol *proto;
 883         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 884         int err = 0;
 885
 886         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 887
 888         proto = ip_conntrack_proto_find_get(npt);
 889
 890         if (proto->from_nfattr)
 891                 err = proto->from_nfattr(tb, ct);
 892         ip_conntrack_proto_put(proto);
 893
 894         return err;
 895 }
 896
 897 static int
 898 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 899 {
 900         int err;
 901
 902         if (cda[CTA_HELP-1]) {
 903                 err = ctnetlink_change_helper(ct, cda);
 904                 if (err < 0)
 905                         return err;
 906         }
 907
 908         if (cda[CTA_TIMEOUT-1]) {
 909                 err = ctnetlink_change_timeout(ct, cda);
 910                 if (err < 0)
 911                         return err;
 912         }
 913
 914         if (cda[CTA_STATUS-1]) {
 915                 err = ctnetlink_change_status(ct, cda);
 916                 if (err < 0)
 917                         return err;
 918         }
 919
 920         if (cda[CTA_PROTOINFO-1]) {
 921                 err = ctnetlink_change_protoinfo(ct, cda);
 922                 if (err < 0)
 923                         return err;
 924         }
 925
 926 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 927         if (cda[CTA_MARK-1])
 928                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 929 #endif
 930
 931         return 0;
 932 }
 933
 934 static int
 935 ctnetlink_create_conntrack(struct nfattr *cda[],
 936                            struct ip_conntrack_tuple *otuple,
 937                            struct ip_conntrack_tuple *rtuple)
 938 {
 939         struct ip_conntrack *ct;
 940         int err = -EINVAL;
 941
 942         ct = ip_conntrack_alloc(otuple, rtuple);
 943         if (ct == NULL || IS_ERR(ct))
 944                 return -ENOMEM;
 945
 946         if (!cda[CTA_TIMEOUT-1])
 947                 goto err;
 948         ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 949
 950         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
 951         ct->status |= IPS_CONFIRMED;
 952
 953         if (cda[CTA_STATUS-1]) {
 954                 err = ctnetlink_change_status(ct, cda);
 955                 if (err < 0)
 956                         goto err;
 957         }
 958
 959         if (cda[CTA_PROTOINFO-1]) {
 960                 err = ctnetlink_change_protoinfo(ct, cda);
 961                 if (err < 0)
 962                         goto err;
 963         }
 964
 965 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 966         if (cda[CTA_MARK-1])
 967                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 968 #endif
 969
 970         ct->helper = ip_conntrack_helper_find_get(rtuple);
 971
 972         add_timer(&ct->timeout);
 973         ip_conntrack_hash_insert(ct);
 974
 975         if (ct->helper)
 976                 ip_conntrack_helper_put(ct->helper);
 977
 978         return 0;
 979
 980 err:
 981         ip_conntrack_free(ct);
 982         return err;
 983 }
 984
 985 static int
 986 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 987                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 988 {
 989         struct ip_conntrack_tuple otuple, rtuple;
 990         struct ip_conntrack_tuple_hash *h = NULL;
 991         int err = 0;
 992
 993         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 994                 return -EINVAL;
 995
 996         if (cda[CTA_TUPLE_ORIG-1]) {
 997                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
 998                 if (err < 0)
 999                         return err;
1000         }
1001
1002         if (cda[CTA_TUPLE_REPLY-1]) {
1003                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1004                 if (err < 0)
1005                         return err;
1006         }
1007
1008         write_lock_bh(&ip_conntrack_lock);
1009         if (cda[CTA_TUPLE_ORIG-1])
1010                 h = __ip_conntrack_find(&otuple, NULL);
1011         else if (cda[CTA_TUPLE_REPLY-1])
1012                 h = __ip_conntrack_find(&rtuple, NULL);
1013
1014         if (h == NULL) {
1015                 write_unlock_bh(&ip_conntrack_lock);
1016                 err = -ENOENT;
1017                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1018                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1019                 return err;
1020         }
1021         /* implicit 'else' */
1022
1023         /* we only allow nat config for new conntracks */
1024         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1025                 err = -EINVAL;
1026                 goto out_unlock;
1027         }
1028
1029         /* We manipulate the conntrack inside the global conntrack table lock,
1030          * so there's no need to increase the refcount */
1031         err = -EEXIST;
1032         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1033                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1034
1035 out_unlock:
1036         write_unlock_bh(&ip_conntrack_lock);
1037         return err;
1038 }
1039
1040 /***********************************************************************
1041  * EXPECT
1042  ***********************************************************************/
1043
1044 static inline int
1045 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1046                          const struct ip_conntrack_tuple *tuple,
1047                          enum ctattr_expect type)
1048 {
1049         struct nfattr *nest_parms = NFA_NEST(skb, type);
1050
1051         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1052                 goto nfattr_failure;
1053
1054         NFA_NEST_END(skb, nest_parms);
1055
1056         return 0;
1057
1058 nfattr_failure:
1059         return -1;
1060 }
1061
1062 static inline int
1063 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1064                         const struct ip_conntrack_tuple *tuple,
1065                         const struct ip_conntrack_tuple *mask)
1066 {
1067         int ret;
1068         struct ip_conntrack_protocol *proto;
1069         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1070
1071         ret = ctnetlink_dump_tuples_ip(skb, mask);
1072         if (unlikely(ret < 0))
1073                 goto nfattr_failure;
1074
1075         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1076         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1077         ip_conntrack_proto_put(proto);
1078         if (unlikely(ret < 0))
1079                 goto nfattr_failure;
1080
1081         NFA_NEST_END(skb, nest_parms);
1082
1083         return 0;
1084
1085 nfattr_failure:
1086         return -1;
1087 }
1088
1089 static inline int
1090 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1091                           const struct ip_conntrack_expect *exp)
1092 {
1093         struct ip_conntrack *master = exp->master;
1094         __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1095         __be32 id = htonl(exp->id);
1096
1097         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1098                 goto nfattr_failure;
1099         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1100                 goto nfattr_failure;
1101         if (ctnetlink_exp_dump_tuple(skb,
1102                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1103                                  CTA_EXPECT_MASTER) < 0)
1104                 goto nfattr_failure;
1105
1106         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1107         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1108
1109         return 0;
1110
1111 nfattr_failure:
1112         return -1;
1113 }
1114
1115 static int
1116 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1117                     int event,
1118                     int nowait,
1119                     const struct ip_conntrack_expect *exp)
1120 {
1121         struct nlmsghdr *nlh;
1122         struct nfgenmsg *nfmsg;
1123         unsigned char *b;
1124
1125         b = skb->tail;
1126
1127         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1128         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1129         nfmsg  = NLMSG_DATA(nlh);
1130
1131         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1132         nfmsg->nfgen_family = AF_INET;
1133         nfmsg->version      = NFNETLINK_V0;
1134         nfmsg->res_id       = 0;
1135
1136         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1137                 goto nfattr_failure;
1138
1139         nlh->nlmsg_len = skb->tail - b;
1140         return skb->len;
1141
1142 nlmsg_failure:
1143 nfattr_failure:
1144         skb_trim(skb, b - skb->data);
1145         return -1;
1146 }
1147
1148 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1149 static int ctnetlink_expect_event(struct notifier_block *this,
1150                                   unsigned long events, void *ptr)
1151 {
1152         struct nlmsghdr *nlh;
1153         struct nfgenmsg *nfmsg;
1154         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1155         struct sk_buff *skb;
1156         unsigned int type;
1157         unsigned char *b;
1158         int flags = 0;
1159
1160         if (events & IPEXP_NEW) {
1161                 type = IPCTNL_MSG_EXP_NEW;
1162                 flags = NLM_F_CREATE|NLM_F_EXCL;
1163         } else
1164                 return NOTIFY_DONE;
1165
1166         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1167                 return NOTIFY_DONE;
1168
1169         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1170         if (!skb)
1171                 return NOTIFY_DONE;
1172
1173         b = skb->tail;
1174
1175         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1176         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1177         nfmsg = NLMSG_DATA(nlh);
1178
1179         nlh->nlmsg_flags    = flags;
1180         nfmsg->nfgen_family = AF_INET;
1181         nfmsg->version      = NFNETLINK_V0;
1182         nfmsg->res_id       = 0;
1183
1184         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1185                 goto nfattr_failure;
1186
1187         nlh->nlmsg_len = skb->tail - b;
1188         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1189         return NOTIFY_DONE;
1190
1191 nlmsg_failure:
1192 nfattr_failure:
1193         kfree_skb(skb);
1194         return NOTIFY_DONE;
1195 }
1196 #endif
1197
1198 static int
1199 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1200 {
1201         struct ip_conntrack_expect *exp = NULL;
1202         struct list_head *i;
1203         u_int32_t *id = (u_int32_t *) &cb->args[0];
1204
1205         read_lock_bh(&ip_conntrack_lock);
1206         list_for_each_prev(i, &ip_conntrack_expect_list) {
1207                 exp = (struct ip_conntrack_expect *) i;
1208                 if (exp->id <= *id)
1209                         continue;
1210                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1211                                             cb->nlh->nlmsg_seq,
1212                                             IPCTNL_MSG_EXP_NEW,
1213                                             1, exp) < 0)
1214                         goto out;
1215                 *id = exp->id;
1216         }
1217 out:
1218         read_unlock_bh(&ip_conntrack_lock);
1219
1220         return skb->len;
1221 }
1222
1223 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1224         [CTA_EXPECT_TIMEOUT-1]          = sizeof(__be32),
1225         [CTA_EXPECT_ID-1]               = sizeof(__be32)
1226 };
1227
1228 static int
1229 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1230                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1231 {
1232         struct ip_conntrack_tuple tuple;
1233         struct ip_conntrack_expect *exp;
1234         struct sk_buff *skb2;
1235         int err = 0;
1236
1237         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1238                 return -EINVAL;
1239
1240         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1241                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1242                 u32 rlen;
1243
1244                 if (msg->nfgen_family != AF_INET)
1245                         return -EAFNOSUPPORT;
1246
1247                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1248                                                 ctnetlink_exp_dump_table,
1249                                                 ctnetlink_done)) != 0)
1250                         return -EINVAL;
1251                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1252                 if (rlen > skb->len)
1253                         rlen = skb->len;
1254                 skb_pull(skb, rlen);
1255                 return 0;
1256         }
1257
1258         if (cda[CTA_EXPECT_MASTER-1])
1259                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1260         else
1261                 return -EINVAL;
1262
1263         if (err < 0)
1264                 return err;
1265
1266         exp = ip_conntrack_expect_find_get(&tuple);
1267         if (!exp)
1268                 return -ENOENT;
1269
1270         if (cda[CTA_EXPECT_ID-1]) {
1271                 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1272                 if (exp->id != ntohl(id)) {
1273                         ip_conntrack_expect_put(exp);
1274                         return -ENOENT;
1275                 }
1276         }
1277
1278         err = -ENOMEM;
1279         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1280         if (!skb2)
1281                 goto out;
1282
1283         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1284                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1285                                       1, exp);
1286         if (err <= 0)
1287                 goto free;
1288
1289         ip_conntrack_expect_put(exp);
1290
1291         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1292
1293 free:
1294         kfree_skb(skb2);
1295 out:
1296         ip_conntrack_expect_put(exp);
1297         return err;
1298 }
1299
1300 static int
1301 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1302                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1303 {
1304         struct ip_conntrack_expect *exp, *tmp;
1305         struct ip_conntrack_tuple tuple;
1306         struct ip_conntrack_helper *h;
1307         int err;
1308
1309         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1310                 return -EINVAL;
1311
1312         if (cda[CTA_EXPECT_TUPLE-1]) {
1313                 /* delete a single expect by tuple */
1314                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1315                 if (err < 0)
1316                         return err;
1317
1318                 /* bump usage count to 2 */
1319                 exp = ip_conntrack_expect_find_get(&tuple);
1320                 if (!exp)
1321                         return -ENOENT;
1322
1323                 if (cda[CTA_EXPECT_ID-1]) {
1324                         __be32 id =
1325                                 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1326                         if (exp->id != ntohl(id)) {
1327                                 ip_conntrack_expect_put(exp);
1328                                 return -ENOENT;
1329                         }
1330                 }
1331
1332                 /* after list removal, usage count == 1 */
1333                 ip_conntrack_unexpect_related(exp);
1334                 /* have to put what we 'get' above.
1335                  * after this line usage count == 0 */
1336                 ip_conntrack_expect_put(exp);
1337         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1338                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1339
1340                 /* delete all expectations for this helper */
1341                 write_lock_bh(&ip_conntrack_lock);
1342                 h = __ip_conntrack_helper_find_byname(name);
1343                 if (!h) {
1344                         write_unlock_bh(&ip_conntrack_lock);
1345                         return -EINVAL;
1346                 }
1347                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1348                                          list) {
1349                         if (exp->master->helper == h
1350                             && del_timer(&exp->timeout)) {
1351                                 ip_ct_unlink_expect(exp);
1352                                 ip_conntrack_expect_put(exp);
1353                         }
1354                 }
1355                 write_unlock_bh(&ip_conntrack_lock);
1356         } else {
1357                 /* This basically means we have to flush everything*/
1358                 write_lock_bh(&ip_conntrack_lock);
1359                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1360                                          list) {
1361                         if (del_timer(&exp->timeout)) {
1362                                 ip_ct_unlink_expect(exp);
1363                                 ip_conntrack_expect_put(exp);
1364                         }
1365                 }
1366                 write_unlock_bh(&ip_conntrack_lock);
1367         }
1368
1369         return 0;
1370 }
1371 static int
1372 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1373 {
1374         return -EOPNOTSUPP;
1375 }
1376
1377 static int
1378 ctnetlink_create_expect(struct nfattr *cda[])
1379 {
1380         struct ip_conntrack_tuple tuple, mask, master_tuple;
1381         struct ip_conntrack_tuple_hash *h = NULL;
1382         struct ip_conntrack_expect *exp;
1383         struct ip_conntrack *ct;
1384         int err = 0;
1385
1386         /* caller guarantees that those three CTA_EXPECT_* exist */
1387         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1388         if (err < 0)
1389                 return err;
1390         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1391         if (err < 0)
1392                 return err;
1393         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1394         if (err < 0)
1395                 return err;
1396
1397         /* Look for master conntrack of this expectation */
1398         h = ip_conntrack_find_get(&master_tuple, NULL);
1399         if (!h)
1400                 return -ENOENT;
1401         ct = tuplehash_to_ctrack(h);
1402
1403         if (!ct->helper) {
1404                 /* such conntrack hasn't got any helper, abort */
1405                 err = -EINVAL;
1406                 goto out;
1407         }
1408
1409         exp = ip_conntrack_expect_alloc(ct);
1410         if (!exp) {
1411                 err = -ENOMEM;
1412                 goto out;
1413         }
1414
1415         exp->expectfn = NULL;
1416         exp->flags = 0;
1417         exp->master = ct;
1418         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1419         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1420
1421         err = ip_conntrack_expect_related(exp);
1422         ip_conntrack_expect_put(exp);
1423
1424 out:
1425         ip_conntrack_put(tuplehash_to_ctrack(h));
1426         return err;
1427 }
1428
1429 static int
1430 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1431                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1432 {
1433         struct ip_conntrack_tuple tuple;
1434         struct ip_conntrack_expect *exp;
1435         int err = 0;
1436
1437         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1438                 return -EINVAL;
1439
1440         if (!cda[CTA_EXPECT_TUPLE-1]
1441             || !cda[CTA_EXPECT_MASK-1]
1442             || !cda[CTA_EXPECT_MASTER-1])
1443                 return -EINVAL;
1444
1445         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1446         if (err < 0)
1447                 return err;
1448
1449         write_lock_bh(&ip_conntrack_lock);
1450         exp = __ip_conntrack_expect_find(&tuple);
1451
1452         if (!exp) {
1453                 write_unlock_bh(&ip_conntrack_lock);
1454                 err = -ENOENT;
1455                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1456                         err = ctnetlink_create_expect(cda);
1457                 return err;
1458         }
1459
1460         err = -EEXIST;
1461         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1462                 err = ctnetlink_change_expect(exp, cda);
1463         write_unlock_bh(&ip_conntrack_lock);
1464
1465         return err;
1466 }
1467
1468 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1469 static struct notifier_block ctnl_notifier = {
1470         .notifier_call  = ctnetlink_conntrack_event,
1471 };
1472
1473 static struct notifier_block ctnl_notifier_exp = {
1474         .notifier_call  = ctnetlink_expect_event,
1475 };
1476 #endif
1477
1478 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1479         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1480                                             .attr_count = CTA_MAX, },
1481         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1482                                             .attr_count = CTA_MAX, },
1483         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1484                                             .attr_count = CTA_MAX, },
1485         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1486                                             .attr_count = CTA_MAX, },
1487 };
1488
1489 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1490         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1491                                             .attr_count = CTA_EXPECT_MAX, },
1492         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1493                                             .attr_count = CTA_EXPECT_MAX, },
1494         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1495                                             .attr_count = CTA_EXPECT_MAX, },
1496 };
1497
1498 static struct nfnetlink_subsystem ctnl_subsys = {
1499         .name                           = "conntrack",
1500         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1501         .cb_count                       = IPCTNL_MSG_MAX,
1502         .cb                             = ctnl_cb,
1503 };
1504
1505 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1506         .name                           = "conntrack_expect",
1507         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1508         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1509         .cb                             = ctnl_exp_cb,
1510 };
1511
1512 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1513 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1514
1515 static int __init ctnetlink_init(void)
1516 {
1517         int ret;
1518
1519         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1520         ret = nfnetlink_subsys_register(&ctnl_subsys);
1521         if (ret < 0) {
1522                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1523                 goto err_out;
1524         }
1525
1526         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1527         if (ret < 0) {
1528                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1529                 goto err_unreg_subsys;
1530         }
1531
1532 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1533         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1534         if (ret < 0) {
1535                 printk("ctnetlink_init: cannot register notifier.\n");
1536                 goto err_unreg_exp_subsys;
1537         }
1538
1539         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1540         if (ret < 0) {
1541                 printk("ctnetlink_init: cannot expect register notifier.\n");
1542                 goto err_unreg_notifier;
1543         }
1544 #endif
1545
1546         return 0;
1547
1548 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1549 err_unreg_notifier:
1550         ip_conntrack_unregister_notifier(&ctnl_notifier);
1551 err_unreg_exp_subsys:
1552         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1553 #endif
1554 err_unreg_subsys:
1555         nfnetlink_subsys_unregister(&ctnl_subsys);
1556 err_out:
1557         return ret;
1558 }
1559
1560 static void __exit ctnetlink_exit(void)
1561 {
1562         printk("ctnetlink: unregistering from nfnetlink.\n");
1563
1564 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1565         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1566         ip_conntrack_unregister_notifier(&ctnl_notifier);
1567 #endif
1568
1569         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1570         nfnetlink_subsys_unregister(&ctnl_subsys);
1571         return;
1572 }
1573
1574 module_init(ctnetlink_init);
1575 module_exit(ctnetlink_exit);