git.oblomov.eu Git - linux-2.6/blob - 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         return -1;
 157 }
 158
 159 static inline int
 160 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 161 {
 162         struct nfattr *nest_helper;
 163
 164         if (!ct->helper)
 165                 return 0;
 166
 167         nest_helper = NFA_NEST(skb, CTA_HELP);
 168         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 169
 170         if (ct->helper->to_nfattr)
 171                 ct->helper->to_nfattr(skb, ct);
 172
 173         NFA_NEST_END(skb, nest_helper);
 174
 175         return 0;
 176
 177 nfattr_failure:
 178         return -1;
 179 }
 180
 181 #ifdef CONFIG_IP_NF_CT_ACCT
 182 static inline int
 183 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 184                         enum ip_conntrack_dir dir)
 185 {
 186         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 187         struct nfattr *nest_count = NFA_NEST(skb, type);
 188         __be32 tmp;
 189
 190         tmp = htonl(ct->counters[dir].packets);
 191         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(__be32), &tmp);
 192
 193         tmp = htonl(ct->counters[dir].bytes);
 194         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(__be32), &tmp);
 195
 196         NFA_NEST_END(skb, nest_count);
 197
 198         return 0;
 199
 200 nfattr_failure:
 201         return -1;
 202 }
 203 #else
 204 #define ctnetlink_dump_counters(a, b, c) (0)
 205 #endif
 206
 207 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 208 static inline int
 209 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 210 {
 211         __be32 mark = htonl(ct->mark);
 212
 213         NFA_PUT(skb, CTA_MARK, sizeof(__be32), &mark);
 214         return 0;
 215
 216 nfattr_failure:
 217         return -1;
 218 }
 219 #else
 220 #define ctnetlink_dump_mark(a, b) (0)
 221 #endif
 222
 223 static inline int
 224 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 225 {
 226         __be32 id = htonl(ct->id);
 227         NFA_PUT(skb, CTA_ID, sizeof(__be32), &id);
 228         return 0;
 229
 230 nfattr_failure:
 231         return -1;
 232 }
 233
 234 static inline int
 235 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 236 {
 237         __be32 use = htonl(atomic_read(&ct->ct_general.use));
 238
 239         NFA_PUT(skb, CTA_USE, sizeof(__be32), &use);
 240         return 0;
 241
 242 nfattr_failure:
 243         return -1;
 244 }
 245
 246 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 247
 248 static int
 249 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 250                     int event, int nowait,
 251                     const struct ip_conntrack *ct)
 252 {
 253         struct nlmsghdr *nlh;
 254         struct nfgenmsg *nfmsg;
 255         struct nfattr *nest_parms;
 256         unsigned char *b;
 257
 258         b = skb->tail;
 259
 260         event |= NFNL_SUBSYS_CTNETLINK << 8;
 261         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 262         nfmsg  = NLMSG_DATA(nlh);
 263
 264         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 265         nfmsg->nfgen_family = AF_INET;
 266         nfmsg->version      = NFNETLINK_V0;
 267         nfmsg->res_id       = 0;
 268
 269         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 270         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 271                 goto nfattr_failure;
 272         NFA_NEST_END(skb, nest_parms);
 273
 274         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 275         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 276                 goto nfattr_failure;
 277         NFA_NEST_END(skb, nest_parms);
 278
 279         if (ctnetlink_dump_status(skb, ct) < 0 ||
 280             ctnetlink_dump_timeout(skb, ct) < 0 ||
 281             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 282             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 283             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 284             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 285             ctnetlink_dump_mark(skb, ct) < 0 ||
 286             ctnetlink_dump_id(skb, ct) < 0 ||
 287             ctnetlink_dump_use(skb, ct) < 0)
 288                 goto nfattr_failure;
 289
 290         nlh->nlmsg_len = skb->tail - b;
 291         return skb->len;
 292
 293 nlmsg_failure:
 294 nfattr_failure:
 295         skb_trim(skb, b - skb->data);
 296         return -1;
 297 }
 298
 299 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 300 static int ctnetlink_conntrack_event(struct notifier_block *this,
 301                                      unsigned long events, void *ptr)
 302 {
 303         struct nlmsghdr *nlh;
 304         struct nfgenmsg *nfmsg;
 305         struct nfattr *nest_parms;
 306         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 307         struct sk_buff *skb;
 308         unsigned int type;
 309         unsigned char *b;
 310         unsigned int flags = 0, group;
 311
 312         /* ignore our fake conntrack entry */
 313         if (ct == &ip_conntrack_untracked)
 314                 return NOTIFY_DONE;
 315
 316         if (events & IPCT_DESTROY) {
 317                 type = IPCTNL_MSG_CT_DELETE;
 318                 group = NFNLGRP_CONNTRACK_DESTROY;
 319         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 320                 type = IPCTNL_MSG_CT_NEW;
 321                 flags = NLM_F_CREATE|NLM_F_EXCL;
 322                 /* dump everything */
 323                 events = ~0UL;
 324                 group = NFNLGRP_CONNTRACK_NEW;
 325         } else if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
 326                 type = IPCTNL_MSG_CT_NEW;
 327                 group = NFNLGRP_CONNTRACK_UPDATE;
 328         } else
 329                 return NOTIFY_DONE;
 330
 331         if (!nfnetlink_has_listeners(group))
 332                 return NOTIFY_DONE;
 333
 334         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 335         if (!skb)
 336                 return NOTIFY_DONE;
 337
 338         b = skb->tail;
 339
 340         type |= NFNL_SUBSYS_CTNETLINK << 8;
 341         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 342         nfmsg = NLMSG_DATA(nlh);
 343
 344         nlh->nlmsg_flags    = flags;
 345         nfmsg->nfgen_family = AF_INET;
 346         nfmsg->version  = NFNETLINK_V0;
 347         nfmsg->res_id   = 0;
 348
 349         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 350         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 351                 goto nfattr_failure;
 352         NFA_NEST_END(skb, nest_parms);
 353
 354         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 355         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 356                 goto nfattr_failure;
 357         NFA_NEST_END(skb, nest_parms);
 358
 359         /* NAT stuff is now a status flag */
 360         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 361             && ctnetlink_dump_status(skb, ct) < 0)
 362                 goto nfattr_failure;
 363         if (events & IPCT_REFRESH
 364             && ctnetlink_dump_timeout(skb, ct) < 0)
 365                 goto nfattr_failure;
 366         if (events & IPCT_PROTOINFO
 367             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 368                 goto nfattr_failure;
 369         if (events & IPCT_HELPINFO
 370             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 371                 goto nfattr_failure;
 372
 373         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 374             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 375                 goto nfattr_failure;
 376
 377         if (events & IPCT_MARK
 378             && ctnetlink_dump_mark(skb, ct) < 0)
 379                 goto nfattr_failure;
 380
 381         nlh->nlmsg_len = skb->tail - b;
 382         nfnetlink_send(skb, 0, group, 0);
 383         return NOTIFY_DONE;
 384
 385 nlmsg_failure:
 386 nfattr_failure:
 387         kfree_skb(skb);
 388         return NOTIFY_DONE;
 389 }
 390 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 391
 392 static int ctnetlink_done(struct netlink_callback *cb)
 393 {
 394         if (cb->args[1])
 395                 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
 396         return 0;
 397 }
 398
 399 static int
 400 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 401 {
 402         struct ip_conntrack *ct, *last;
 403         struct ip_conntrack_tuple_hash *h;
 404         struct list_head *i;
 405
 406         read_lock_bh(&ip_conntrack_lock);
 407         last = (struct ip_conntrack *)cb->args[1];
 408         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
 409 restart:
 410                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 411                         h = (struct ip_conntrack_tuple_hash *) i;
 412                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 413                                 continue;
 414                         ct = tuplehash_to_ctrack(h);
 415                         if (cb->args[1]) {
 416                                 if (ct != last)
 417                                         continue;
 418                                 cb->args[1] = 0;
 419                         }
 420                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 421                                                 cb->nlh->nlmsg_seq,
 422                                                 IPCTNL_MSG_CT_NEW,
 423                                                 1, ct) < 0) {
 424                                 nf_conntrack_get(&ct->ct_general);
 425                                 cb->args[1] = (unsigned long)ct;
 426                                 goto out;
 427                         }
 428 #ifdef CONFIG_NF_CT_ACCT
 429                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
 430                                                 IPCTNL_MSG_CT_GET_CTRZERO)
 431                                 memset(&ct->counters, 0, sizeof(ct->counters));
 432 #endif
 433                 }
 434                 if (cb->args[1]) {
 435                         cb->args[1] = 0;
 436                         goto restart;
 437                 }
 438         }
 439 out:
 440         read_unlock_bh(&ip_conntrack_lock);
 441         if (last)
 442                 ip_conntrack_put(last);
 443
 444         return skb->len;
 445 }
 446
 447 static const size_t cta_min_ip[CTA_IP_MAX] = {
 448         [CTA_IP_V4_SRC-1]       = sizeof(__be32),
 449         [CTA_IP_V4_DST-1]       = sizeof(__be32),
 450 };
 451
 452 static inline int
 453 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 454 {
 455         struct nfattr *tb[CTA_IP_MAX];
 456
 457         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 458
 459         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 460                 return -EINVAL;
 461
 462         if (!tb[CTA_IP_V4_SRC-1])
 463                 return -EINVAL;
 464         tuple->src.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 465
 466         if (!tb[CTA_IP_V4_DST-1])
 467                 return -EINVAL;
 468         tuple->dst.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 469
 470         return 0;
 471 }
 472
 473 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 474         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 475         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 476         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 477         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 478         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 479         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 480 };
 481
 482 static inline int
 483 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 484                             struct ip_conntrack_tuple *tuple)
 485 {
 486         struct nfattr *tb[CTA_PROTO_MAX];
 487         struct ip_conntrack_protocol *proto;
 488         int ret = 0;
 489
 490         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 491
 492         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 493                 return -EINVAL;
 494
 495         if (!tb[CTA_PROTO_NUM-1])
 496                 return -EINVAL;
 497         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 498
 499         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 500
 501         if (likely(proto->nfattr_to_tuple))
 502                 ret = proto->nfattr_to_tuple(tb, tuple);
 503
 504         ip_conntrack_proto_put(proto);
 505
 506         return ret;
 507 }
 508
 509 static inline int
 510 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 511                       enum ctattr_tuple type)
 512 {
 513         struct nfattr *tb[CTA_TUPLE_MAX];
 514         int err;
 515
 516         memset(tuple, 0, sizeof(*tuple));
 517
 518         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 519
 520         if (!tb[CTA_TUPLE_IP-1])
 521                 return -EINVAL;
 522
 523         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 524         if (err < 0)
 525                 return err;
 526
 527         if (!tb[CTA_TUPLE_PROTO-1])
 528                 return -EINVAL;
 529
 530         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 531         if (err < 0)
 532                 return err;
 533
 534         /* orig and expect tuples get DIR_ORIGINAL */
 535         if (type == CTA_TUPLE_REPLY)
 536                 tuple->dst.dir = IP_CT_DIR_REPLY;
 537         else
 538                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 539
 540         return 0;
 541 }
 542
 543 #ifdef CONFIG_IP_NF_NAT_NEEDED
 544 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 545         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 546         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 547 };
 548
 549 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 550                                      const struct ip_conntrack *ct,
 551                                      struct ip_nat_range *range)
 552 {
 553         struct nfattr *tb[CTA_PROTONAT_MAX];
 554         struct ip_nat_protocol *npt;
 555
 556         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 557
 558         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 559                 return -EINVAL;
 560
 561         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 562
 563         if (!npt->nfattr_to_range) {
 564                 ip_nat_proto_put(npt);
 565                 return 0;
 566         }
 567
 568         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 569         if (npt->nfattr_to_range(tb, range) > 0)
 570                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 571
 572         ip_nat_proto_put(npt);
 573
 574         return 0;
 575 }
 576
 577 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 578         [CTA_NAT_MINIP-1]       = sizeof(__be32),
 579         [CTA_NAT_MAXIP-1]       = sizeof(__be32),
 580 };
 581
 582 static inline int
 583 ctnetlink_parse_nat(struct nfattr *nat,
 584                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 585 {
 586         struct nfattr *tb[CTA_NAT_MAX];
 587         int err;
 588
 589         memset(range, 0, sizeof(*range));
 590
 591         nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
 592
 593         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 594                 return -EINVAL;
 595
 596         if (tb[CTA_NAT_MINIP-1])
 597                 range->min_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 598
 599         if (!tb[CTA_NAT_MAXIP-1])
 600                 range->max_ip = range->min_ip;
 601         else
 602                 range->max_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 603
 604         if (range->min_ip)
 605                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 606
 607         if (!tb[CTA_NAT_PROTO-1])
 608                 return 0;
 609
 610         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 611         if (err < 0)
 612                 return err;
 613
 614         return 0;
 615 }
 616 #endif
 617
 618 static inline int
 619 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 620 {
 621         struct nfattr *tb[CTA_HELP_MAX];
 622
 623         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 624
 625         if (!tb[CTA_HELP_NAME-1])
 626                 return -EINVAL;
 627
 628         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 629
 630         return 0;
 631 }
 632
 633 static const size_t cta_min[CTA_MAX] = {
 634         [CTA_STATUS-1]          = sizeof(__be32),
 635         [CTA_TIMEOUT-1]         = sizeof(__be32),
 636         [CTA_MARK-1]            = sizeof(__be32),
 637         [CTA_USE-1]             = sizeof(__be32),
 638         [CTA_ID-1]              = sizeof(__be32)
 639 };
 640
 641 static int
 642 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 643                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 644 {
 645         struct ip_conntrack_tuple_hash *h;
 646         struct ip_conntrack_tuple tuple;
 647         struct ip_conntrack *ct;
 648         int err = 0;
 649
 650         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 651                 return -EINVAL;
 652
 653         if (cda[CTA_TUPLE_ORIG-1])
 654                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 655         else if (cda[CTA_TUPLE_REPLY-1])
 656                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 657         else {
 658                 /* Flush the whole table */
 659                 ip_conntrack_flush();
 660                 return 0;
 661         }
 662
 663         if (err < 0)
 664                 return err;
 665
 666         h = ip_conntrack_find_get(&tuple, NULL);
 667         if (!h)
 668                 return -ENOENT;
 669
 670         ct = tuplehash_to_ctrack(h);
 671
 672         if (cda[CTA_ID-1]) {
 673                 u_int32_t id = ntohl(*(__be32 *)NFA_DATA(cda[CTA_ID-1]));
 674                 if (ct->id != id) {
 675                         ip_conntrack_put(ct);
 676                         return -ENOENT;
 677                 }
 678         }
 679         if (del_timer(&ct->timeout))
 680                 ct->timeout.function((unsigned long)ct);
 681
 682         ip_conntrack_put(ct);
 683
 684         return 0;
 685 }
 686
 687 static int
 688 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 689                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 690 {
 691         struct ip_conntrack_tuple_hash *h;
 692         struct ip_conntrack_tuple tuple;
 693         struct ip_conntrack *ct;
 694         struct sk_buff *skb2 = NULL;
 695         int err = 0;
 696
 697         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 698                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 699                 u32 rlen;
 700
 701                 if (msg->nfgen_family != AF_INET)
 702                         return -EAFNOSUPPORT;
 703
 704 #ifndef CONFIG_IP_NF_CT_ACCT
 705                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
 706                         return -ENOTSUPP;
 707 #endif
 708                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 709                                                 ctnetlink_dump_table,
 710                                                 ctnetlink_done)) != 0)
 711                         return -EINVAL;
 712
 713                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 714                 if (rlen > skb->len)
 715                         rlen = skb->len;
 716                 skb_pull(skb, rlen);
 717                 return 0;
 718         }
 719
 720         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 721                 return -EINVAL;
 722
 723         if (cda[CTA_TUPLE_ORIG-1])
 724                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 725         else if (cda[CTA_TUPLE_REPLY-1])
 726                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 727         else
 728                 return -EINVAL;
 729
 730         if (err < 0)
 731                 return err;
 732
 733         h = ip_conntrack_find_get(&tuple, NULL);
 734         if (!h)
 735                 return -ENOENT;
 736
 737         ct = tuplehash_to_ctrack(h);
 738
 739         err = -ENOMEM;
 740         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 741         if (!skb2) {
 742                 ip_conntrack_put(ct);
 743                 return -ENOMEM;
 744         }
 745         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 746
 747         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 748                                   IPCTNL_MSG_CT_NEW, 1, ct);
 749         ip_conntrack_put(ct);
 750         if (err <= 0)
 751                 goto free;
 752
 753         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 754         if (err < 0)
 755                 goto out;
 756
 757         return 0;
 758
 759 free:
 760         kfree_skb(skb2);
 761 out:
 762         return err;
 763 }
 764
 765 static inline int
 766 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 767 {
 768         unsigned long d;
 769         unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
 770         d = ct->status ^ status;
 771
 772         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 773                 /* unchangeable */
 774                 return -EINVAL;
 775
 776         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 777                 /* SEEN_REPLY bit can only be set */
 778                 return -EINVAL;
 779
 780
 781         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 782                 /* ASSURED bit can only be set */
 783                 return -EINVAL;
 784
 785         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 786 #ifndef CONFIG_IP_NF_NAT_NEEDED
 787                 return -EINVAL;
 788 #else
 789                 struct ip_nat_range range;
 790
 791                 if (cda[CTA_NAT_DST-1]) {
 792                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 793                                                 &range) < 0)
 794                                 return -EINVAL;
 795                         if (ip_nat_initialized(ct,
 796                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 797                                 return -EEXIST;
 798                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 799                 }
 800                 if (cda[CTA_NAT_SRC-1]) {
 801                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 802                                                 &range) < 0)
 803                                 return -EINVAL;
 804                         if (ip_nat_initialized(ct,
 805                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 806                                 return -EEXIST;
 807                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 808                 }
 809 #endif
 810         }
 811
 812         /* Be careful here, modifying NAT bits can screw up things,
 813          * so don't let users modify them directly if they don't pass
 814          * ip_nat_range. */
 815         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 816         return 0;
 817 }
 818
 819
 820 static inline int
 821 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 822 {
 823         struct ip_conntrack_helper *helper;
 824         char *helpname;
 825         int err;
 826
 827         /* don't change helper of sibling connections */
 828         if (ct->master)
 829                 return -EINVAL;
 830
 831         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 832         if (err < 0)
 833                 return err;
 834
 835         helper = __ip_conntrack_helper_find_byname(helpname);
 836         if (!helper) {
 837                 if (!strcmp(helpname, ""))
 838                         helper = NULL;
 839                 else
 840                         return -EINVAL;
 841         }
 842
 843         if (ct->helper) {
 844                 if (!helper) {
 845                         /* we had a helper before ... */
 846                         ip_ct_remove_expectations(ct);
 847                         ct->helper = NULL;
 848                 } else {
 849                         /* need to zero data of old helper */
 850                         memset(&ct->help, 0, sizeof(ct->help));
 851                 }
 852         }
 853
 854         ct->helper = helper;
 855
 856         return 0;
 857 }
 858
 859 static inline int
 860 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 861 {
 862         u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 863
 864         if (!del_timer(&ct->timeout))
 865                 return -ETIME;
 866
 867         ct->timeout.expires = jiffies + timeout * HZ;
 868         add_timer(&ct->timeout);
 869
 870         return 0;
 871 }
 872
 873 static inline int
 874 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 875 {
 876         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 877         struct ip_conntrack_protocol *proto;
 878         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 879         int err = 0;
 880
 881         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 882
 883         proto = ip_conntrack_proto_find_get(npt);
 884
 885         if (proto->from_nfattr)
 886                 err = proto->from_nfattr(tb, ct);
 887         ip_conntrack_proto_put(proto);
 888
 889         return err;
 890 }
 891
 892 static int
 893 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 894 {
 895         int err;
 896
 897         if (cda[CTA_HELP-1]) {
 898                 err = ctnetlink_change_helper(ct, cda);
 899                 if (err < 0)
 900                         return err;
 901         }
 902
 903         if (cda[CTA_TIMEOUT-1]) {
 904                 err = ctnetlink_change_timeout(ct, cda);
 905                 if (err < 0)
 906                         return err;
 907         }
 908
 909         if (cda[CTA_STATUS-1]) {
 910                 err = ctnetlink_change_status(ct, cda);
 911                 if (err < 0)
 912                         return err;
 913         }
 914
 915         if (cda[CTA_PROTOINFO-1]) {
 916                 err = ctnetlink_change_protoinfo(ct, cda);
 917                 if (err < 0)
 918                         return err;
 919         }
 920
 921 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 922         if (cda[CTA_MARK-1])
 923                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 924 #endif
 925
 926         return 0;
 927 }
 928
 929 static int
 930 ctnetlink_create_conntrack(struct nfattr *cda[],
 931                            struct ip_conntrack_tuple *otuple,
 932                            struct ip_conntrack_tuple *rtuple)
 933 {
 934         struct ip_conntrack *ct;
 935         int err = -EINVAL;
 936
 937         ct = ip_conntrack_alloc(otuple, rtuple);
 938         if (ct == NULL || IS_ERR(ct))
 939                 return -ENOMEM;
 940
 941         if (!cda[CTA_TIMEOUT-1])
 942                 goto err;
 943         ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 944
 945         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
 946         ct->status |= IPS_CONFIRMED;
 947
 948         err = ctnetlink_change_status(ct, cda);
 949         if (err < 0)
 950                 goto err;
 951
 952         if (cda[CTA_PROTOINFO-1]) {
 953                 err = ctnetlink_change_protoinfo(ct, cda);
 954                 if (err < 0)
 955                         return err;
 956         }
 957
 958 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 959         if (cda[CTA_MARK-1])
 960                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 961 #endif
 962
 963         ct->helper = ip_conntrack_helper_find_get(rtuple);
 964
 965         add_timer(&ct->timeout);
 966         ip_conntrack_hash_insert(ct);
 967
 968         if (ct->helper)
 969                 ip_conntrack_helper_put(ct->helper);
 970
 971         return 0;
 972
 973 err:
 974         ip_conntrack_free(ct);
 975         return err;
 976 }
 977
 978 static int
 979 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 980                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 981 {
 982         struct ip_conntrack_tuple otuple, rtuple;
 983         struct ip_conntrack_tuple_hash *h = NULL;
 984         int err = 0;
 985
 986         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 987                 return -EINVAL;
 988
 989         if (cda[CTA_TUPLE_ORIG-1]) {
 990                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
 991                 if (err < 0)
 992                         return err;
 993         }
 994
 995         if (cda[CTA_TUPLE_REPLY-1]) {
 996                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
 997                 if (err < 0)
 998                         return err;
 999         }
1000
1001         write_lock_bh(&ip_conntrack_lock);
1002         if (cda[CTA_TUPLE_ORIG-1])
1003                 h = __ip_conntrack_find(&otuple, NULL);
1004         else if (cda[CTA_TUPLE_REPLY-1])
1005                 h = __ip_conntrack_find(&rtuple, NULL);
1006
1007         if (h == NULL) {
1008                 write_unlock_bh(&ip_conntrack_lock);
1009                 err = -ENOENT;
1010                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1011                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1012                 return err;
1013         }
1014         /* implicit 'else' */
1015
1016         /* we only allow nat config for new conntracks */
1017         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1018                 err = -EINVAL;
1019                 goto out_unlock;
1020         }
1021
1022         /* We manipulate the conntrack inside the global conntrack table lock,
1023          * so there's no need to increase the refcount */
1024         err = -EEXIST;
1025         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1026                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1027
1028 out_unlock:
1029         write_unlock_bh(&ip_conntrack_lock);
1030         return err;
1031 }
1032
1033 /***********************************************************************
1034  * EXPECT
1035  ***********************************************************************/
1036
1037 static inline int
1038 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1039                          const struct ip_conntrack_tuple *tuple,
1040                          enum ctattr_expect type)
1041 {
1042         struct nfattr *nest_parms = NFA_NEST(skb, type);
1043
1044         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1045                 goto nfattr_failure;
1046
1047         NFA_NEST_END(skb, nest_parms);
1048
1049         return 0;
1050
1051 nfattr_failure:
1052         return -1;
1053 }
1054
1055 static inline int
1056 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1057                         const struct ip_conntrack_tuple *tuple,
1058                         const struct ip_conntrack_tuple *mask)
1059 {
1060         int ret;
1061         struct ip_conntrack_protocol *proto;
1062         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1063
1064         ret = ctnetlink_dump_tuples_ip(skb, mask);
1065         if (unlikely(ret < 0))
1066                 goto nfattr_failure;
1067
1068         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1069         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1070         ip_conntrack_proto_put(proto);
1071         if (unlikely(ret < 0))
1072                 goto nfattr_failure;
1073
1074         NFA_NEST_END(skb, nest_parms);
1075
1076         return 0;
1077
1078 nfattr_failure:
1079         return -1;
1080 }
1081
1082 static inline int
1083 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1084                           const struct ip_conntrack_expect *exp)
1085 {
1086         struct ip_conntrack *master = exp->master;
1087         __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1088         __be32 id = htonl(exp->id);
1089
1090         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1091                 goto nfattr_failure;
1092         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1093                 goto nfattr_failure;
1094         if (ctnetlink_exp_dump_tuple(skb,
1095                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1096                                  CTA_EXPECT_MASTER) < 0)
1097                 goto nfattr_failure;
1098
1099         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1100         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1101
1102         return 0;
1103
1104 nfattr_failure:
1105         return -1;
1106 }
1107
1108 static int
1109 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1110                     int event,
1111                     int nowait,
1112                     const struct ip_conntrack_expect *exp)
1113 {
1114         struct nlmsghdr *nlh;
1115         struct nfgenmsg *nfmsg;
1116         unsigned char *b;
1117
1118         b = skb->tail;
1119
1120         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1121         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1122         nfmsg  = NLMSG_DATA(nlh);
1123
1124         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1125         nfmsg->nfgen_family = AF_INET;
1126         nfmsg->version      = NFNETLINK_V0;
1127         nfmsg->res_id       = 0;
1128
1129         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1130                 goto nfattr_failure;
1131
1132         nlh->nlmsg_len = skb->tail - b;
1133         return skb->len;
1134
1135 nlmsg_failure:
1136 nfattr_failure:
1137         skb_trim(skb, b - skb->data);
1138         return -1;
1139 }
1140
1141 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1142 static int ctnetlink_expect_event(struct notifier_block *this,
1143                                   unsigned long events, void *ptr)
1144 {
1145         struct nlmsghdr *nlh;
1146         struct nfgenmsg *nfmsg;
1147         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1148         struct sk_buff *skb;
1149         unsigned int type;
1150         unsigned char *b;
1151         int flags = 0;
1152
1153         if (events & IPEXP_NEW) {
1154                 type = IPCTNL_MSG_EXP_NEW;
1155                 flags = NLM_F_CREATE|NLM_F_EXCL;
1156         } else
1157                 return NOTIFY_DONE;
1158
1159         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1160                 return NOTIFY_DONE;
1161
1162         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1163         if (!skb)
1164                 return NOTIFY_DONE;
1165
1166         b = skb->tail;
1167
1168         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1169         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1170         nfmsg = NLMSG_DATA(nlh);
1171
1172         nlh->nlmsg_flags    = flags;
1173         nfmsg->nfgen_family = AF_INET;
1174         nfmsg->version      = NFNETLINK_V0;
1175         nfmsg->res_id       = 0;
1176
1177         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1178                 goto nfattr_failure;
1179
1180         nlh->nlmsg_len = skb->tail - b;
1181         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1182         return NOTIFY_DONE;
1183
1184 nlmsg_failure:
1185 nfattr_failure:
1186         kfree_skb(skb);
1187         return NOTIFY_DONE;
1188 }
1189 #endif
1190
1191 static int
1192 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1193 {
1194         struct ip_conntrack_expect *exp = NULL;
1195         struct list_head *i;
1196         u_int32_t *id = (u_int32_t *) &cb->args[0];
1197
1198         read_lock_bh(&ip_conntrack_lock);
1199         list_for_each_prev(i, &ip_conntrack_expect_list) {
1200                 exp = (struct ip_conntrack_expect *) i;
1201                 if (exp->id <= *id)
1202                         continue;
1203                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1204                                             cb->nlh->nlmsg_seq,
1205                                             IPCTNL_MSG_EXP_NEW,
1206                                             1, exp) < 0)
1207                         goto out;
1208                 *id = exp->id;
1209         }
1210 out:
1211         read_unlock_bh(&ip_conntrack_lock);
1212
1213         return skb->len;
1214 }
1215
1216 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1217         [CTA_EXPECT_TIMEOUT-1]          = sizeof(__be32),
1218         [CTA_EXPECT_ID-1]               = sizeof(__be32)
1219 };
1220
1221 static int
1222 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1223                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1224 {
1225         struct ip_conntrack_tuple tuple;
1226         struct ip_conntrack_expect *exp;
1227         struct sk_buff *skb2;
1228         int err = 0;
1229
1230         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1231                 return -EINVAL;
1232
1233         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1234                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1235                 u32 rlen;
1236
1237                 if (msg->nfgen_family != AF_INET)
1238                         return -EAFNOSUPPORT;
1239
1240                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1241                                                 ctnetlink_exp_dump_table,
1242                                                 ctnetlink_done)) != 0)
1243                         return -EINVAL;
1244                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1245                 if (rlen > skb->len)
1246                         rlen = skb->len;
1247                 skb_pull(skb, rlen);
1248                 return 0;
1249         }
1250
1251         if (cda[CTA_EXPECT_MASTER-1])
1252                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1253         else
1254                 return -EINVAL;
1255
1256         if (err < 0)
1257                 return err;
1258
1259         exp = ip_conntrack_expect_find(&tuple);
1260         if (!exp)
1261                 return -ENOENT;
1262
1263         if (cda[CTA_EXPECT_ID-1]) {
1264                 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1265                 if (exp->id != ntohl(id)) {
1266                         ip_conntrack_expect_put(exp);
1267                         return -ENOENT;
1268                 }
1269         }
1270
1271         err = -ENOMEM;
1272         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1273         if (!skb2)
1274                 goto out;
1275         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1276
1277         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1278                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1279                                       1, exp);
1280         if (err <= 0)
1281                 goto free;
1282
1283         ip_conntrack_expect_put(exp);
1284
1285         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1286
1287 free:
1288         kfree_skb(skb2);
1289 out:
1290         ip_conntrack_expect_put(exp);
1291         return err;
1292 }
1293
1294 static int
1295 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1296                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1297 {
1298         struct ip_conntrack_expect *exp, *tmp;
1299         struct ip_conntrack_tuple tuple;
1300         struct ip_conntrack_helper *h;
1301         int err;
1302
1303         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1304                 return -EINVAL;
1305
1306         if (cda[CTA_EXPECT_TUPLE-1]) {
1307                 /* delete a single expect by tuple */
1308                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1309                 if (err < 0)
1310                         return err;
1311
1312                 /* bump usage count to 2 */
1313                 exp = ip_conntrack_expect_find(&tuple);
1314                 if (!exp)
1315                         return -ENOENT;
1316
1317                 if (cda[CTA_EXPECT_ID-1]) {
1318                         __be32 id =
1319                                 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1320                         if (exp->id != ntohl(id)) {
1321                                 ip_conntrack_expect_put(exp);
1322                                 return -ENOENT;
1323                         }
1324                 }
1325
1326                 /* after list removal, usage count == 1 */
1327                 ip_conntrack_unexpect_related(exp);
1328                 /* have to put what we 'get' above.
1329                  * after this line usage count == 0 */
1330                 ip_conntrack_expect_put(exp);
1331         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1332                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1333
1334                 /* delete all expectations for this helper */
1335                 write_lock_bh(&ip_conntrack_lock);
1336                 h = __ip_conntrack_helper_find_byname(name);
1337                 if (!h) {
1338                         write_unlock_bh(&ip_conntrack_lock);
1339                         return -EINVAL;
1340                 }
1341                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1342                                          list) {
1343                         if (exp->master->helper == h
1344                             && del_timer(&exp->timeout)) {
1345                                 ip_ct_unlink_expect(exp);
1346                                 ip_conntrack_expect_put(exp);
1347                         }
1348                 }
1349                 write_unlock_bh(&ip_conntrack_lock);
1350         } else {
1351                 /* This basically means we have to flush everything*/
1352                 write_lock_bh(&ip_conntrack_lock);
1353                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1354                                          list) {
1355                         if (del_timer(&exp->timeout)) {
1356                                 ip_ct_unlink_expect(exp);
1357                                 ip_conntrack_expect_put(exp);
1358                         }
1359                 }
1360                 write_unlock_bh(&ip_conntrack_lock);
1361         }
1362
1363         return 0;
1364 }
1365 static int
1366 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1367 {
1368         return -EOPNOTSUPP;
1369 }
1370
1371 static int
1372 ctnetlink_create_expect(struct nfattr *cda[])
1373 {
1374         struct ip_conntrack_tuple tuple, mask, master_tuple;
1375         struct ip_conntrack_tuple_hash *h = NULL;
1376         struct ip_conntrack_expect *exp;
1377         struct ip_conntrack *ct;
1378         int err = 0;
1379
1380         /* caller guarantees that those three CTA_EXPECT_* exist */
1381         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1382         if (err < 0)
1383                 return err;
1384         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1385         if (err < 0)
1386                 return err;
1387         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1388         if (err < 0)
1389                 return err;
1390
1391         /* Look for master conntrack of this expectation */
1392         h = ip_conntrack_find_get(&master_tuple, NULL);
1393         if (!h)
1394                 return -ENOENT;
1395         ct = tuplehash_to_ctrack(h);
1396
1397         if (!ct->helper) {
1398                 /* such conntrack hasn't got any helper, abort */
1399                 err = -EINVAL;
1400                 goto out;
1401         }
1402
1403         exp = ip_conntrack_expect_alloc(ct);
1404         if (!exp) {
1405                 err = -ENOMEM;
1406                 goto out;
1407         }
1408
1409         exp->expectfn = NULL;
1410         exp->flags = 0;
1411         exp->master = ct;
1412         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1413         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1414
1415         err = ip_conntrack_expect_related(exp);
1416         ip_conntrack_expect_put(exp);
1417
1418 out:
1419         ip_conntrack_put(tuplehash_to_ctrack(h));
1420         return err;
1421 }
1422
1423 static int
1424 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1425                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1426 {
1427         struct ip_conntrack_tuple tuple;
1428         struct ip_conntrack_expect *exp;
1429         int err = 0;
1430
1431         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1432                 return -EINVAL;
1433
1434         if (!cda[CTA_EXPECT_TUPLE-1]
1435             || !cda[CTA_EXPECT_MASK-1]
1436             || !cda[CTA_EXPECT_MASTER-1])
1437                 return -EINVAL;
1438
1439         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1440         if (err < 0)
1441                 return err;
1442
1443         write_lock_bh(&ip_conntrack_lock);
1444         exp = __ip_conntrack_expect_find(&tuple);
1445
1446         if (!exp) {
1447                 write_unlock_bh(&ip_conntrack_lock);
1448                 err = -ENOENT;
1449                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1450                         err = ctnetlink_create_expect(cda);
1451                 return err;
1452         }
1453
1454         err = -EEXIST;
1455         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1456                 err = ctnetlink_change_expect(exp, cda);
1457         write_unlock_bh(&ip_conntrack_lock);
1458
1459         return err;
1460 }
1461
1462 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1463 static struct notifier_block ctnl_notifier = {
1464         .notifier_call  = ctnetlink_conntrack_event,
1465 };
1466
1467 static struct notifier_block ctnl_notifier_exp = {
1468         .notifier_call  = ctnetlink_expect_event,
1469 };
1470 #endif
1471
1472 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1473         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1474                                             .attr_count = CTA_MAX, },
1475         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1476                                             .attr_count = CTA_MAX, },
1477         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1478                                             .attr_count = CTA_MAX, },
1479         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1480                                             .attr_count = CTA_MAX, },
1481 };
1482
1483 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1484         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1485                                             .attr_count = CTA_EXPECT_MAX, },
1486         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1487                                             .attr_count = CTA_EXPECT_MAX, },
1488         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1489                                             .attr_count = CTA_EXPECT_MAX, },
1490 };
1491
1492 static struct nfnetlink_subsystem ctnl_subsys = {
1493         .name                           = "conntrack",
1494         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1495         .cb_count                       = IPCTNL_MSG_MAX,
1496         .cb                             = ctnl_cb,
1497 };
1498
1499 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1500         .name                           = "conntrack_expect",
1501         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1502         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1503         .cb                             = ctnl_exp_cb,
1504 };
1505
1506 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1507 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1508
1509 static int __init ctnetlink_init(void)
1510 {
1511         int ret;
1512
1513         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1514         ret = nfnetlink_subsys_register(&ctnl_subsys);
1515         if (ret < 0) {
1516                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1517                 goto err_out;
1518         }
1519
1520         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1521         if (ret < 0) {
1522                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1523                 goto err_unreg_subsys;
1524         }
1525
1526 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1527         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1528         if (ret < 0) {
1529                 printk("ctnetlink_init: cannot register notifier.\n");
1530                 goto err_unreg_exp_subsys;
1531         }
1532
1533         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1534         if (ret < 0) {
1535                 printk("ctnetlink_init: cannot expect register notifier.\n");
1536                 goto err_unreg_notifier;
1537         }
1538 #endif
1539
1540         return 0;
1541
1542 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1543 err_unreg_notifier:
1544         ip_conntrack_unregister_notifier(&ctnl_notifier);
1545 err_unreg_exp_subsys:
1546         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1547 #endif
1548 err_unreg_subsys:
1549         nfnetlink_subsys_unregister(&ctnl_subsys);
1550 err_out:
1551         return ret;
1552 }
1553
1554 static void __exit ctnetlink_exit(void)
1555 {
1556         printk("ctnetlink: unregistering from nfnetlink.\n");
1557
1558 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1559         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1560         ip_conntrack_unregister_notifier(&ctnl_notifier);
1561 #endif
1562
1563         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1564         nfnetlink_subsys_unregister(&ctnl_subsys);
1565         return;
1566 }
1567
1568 module_init(ctnetlink_init);
1569 module_exit(ctnetlink_exit);