SAFE public projects git trees. - safe/jmp/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 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/notifier.h>
  31 #include <linux/rtnetlink.h>
  32
  33 #include <linux/netfilter.h>
  34 #include <linux/netfilter_ipv4.h>
  35 #include <linux/netfilter_ipv4/ip_tables.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  38 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  39 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  40 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  41
  42 #include <linux/netfilter/nfnetlink.h>
  43 #include <linux/netfilter/nfnetlink_conntrack.h>
  44
  45 MODULE_LICENSE("GPL");
  46
  47 static char __initdata version[] = "0.90";
  48
  49 #if 0
  50 #define DEBUGP printk
  51 #else
  52 #define DEBUGP(format, args...)
  53 #endif
  54
  55
  56 static inline int
  57 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  58                             const struct ip_conntrack_tuple *tuple)
  59 {
  60         struct ip_conntrack_protocol *proto;
  61
  62         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  63
  64         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  65         if (proto && proto->tuple_to_nfattr)
  66                 return proto->tuple_to_nfattr(skb, tuple);
  67
  68         return 0;
  69
  70 nfattr_failure:
  71         return -1;
  72 }
  73
  74 static inline int
  75 ctnetlink_dump_tuples(struct sk_buff *skb,
  76                       const struct ip_conntrack_tuple *tuple)
  77 {
  78         struct nfattr *nest_parms;
  79
  80         nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  81         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  82         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  83         NFA_NEST_END(skb, nest_parms);
  84
  85         nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  86         ctnetlink_dump_tuples_proto(skb, tuple);
  87         NFA_NEST_END(skb, nest_parms);
  88
  89         return 0;
  90
  91 nfattr_failure:
  92         return -1;
  93 }
  94
  95 static inline int
  96 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
  97 {
  98         u_int32_t status = htonl((u_int32_t) ct->status);
  99         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 100         return 0;
 101
 102 nfattr_failure:
 103         return -1;
 104 }
 105
 106 static inline int
 107 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 108 {
 109         long timeout_l = ct->timeout.expires - jiffies;
 110         u_int32_t timeout;
 111
 112         if (timeout_l < 0)
 113                 timeout = 0;
 114         else
 115                 timeout = htonl(timeout_l / HZ);
 116
 117         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 118         return 0;
 119
 120 nfattr_failure:
 121         return -1;
 122 }
 123
 124 static inline int
 125 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 126 {
 127         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 128
 129         struct nfattr *nest_proto;
 130         int ret;
 131
 132         if (!proto || !proto->to_nfattr)
 133                 return 0;
 134
 135         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 136
 137         ret = proto->to_nfattr(skb, nest_proto, ct);
 138
 139         ip_conntrack_proto_put(proto);
 140
 141         NFA_NEST_END(skb, nest_proto);
 142
 143         return ret;
 144
 145 nfattr_failure:
 146         return -1;
 147 }
 148
 149 static inline int
 150 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 151 {
 152         struct nfattr *nest_helper;
 153
 154         if (!ct->helper)
 155                 return 0;
 156
 157         nest_helper = NFA_NEST(skb, CTA_HELP);
 158         NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
 159
 160         if (ct->helper->to_nfattr)
 161                 ct->helper->to_nfattr(skb, ct);
 162
 163         NFA_NEST_END(skb, nest_helper);
 164
 165         return 0;
 166
 167 nfattr_failure:
 168         return -1;
 169 }
 170
 171 #ifdef CONFIG_IP_NF_CT_ACCT
 172 static inline int
 173 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 174                         enum ip_conntrack_dir dir)
 175 {
 176         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 177         struct nfattr *nest_count = NFA_NEST(skb, type);
 178         u_int64_t tmp;
 179
 180         tmp = cpu_to_be64(ct->counters[dir].packets);
 181         NFA_PUT(skb, CTA_COUNTERS_PACKETS, sizeof(u_int64_t), &tmp);
 182
 183         tmp = cpu_to_be64(ct->counters[dir].bytes);
 184         NFA_PUT(skb, CTA_COUNTERS_BYTES, sizeof(u_int64_t), &tmp);
 185
 186         NFA_NEST_END(skb, nest_count);
 187
 188         return 0;
 189
 190 nfattr_failure:
 191         return -1;
 192 }
 193 #else
 194 #define ctnetlink_dump_counters(a, b, c) (0)
 195 #endif
 196
 197 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 198 static inline int
 199 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 200 {
 201         u_int32_t mark = htonl(ct->mark);
 202
 203         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 204         return 0;
 205
 206 nfattr_failure:
 207         return -1;
 208 }
 209 #else
 210 #define ctnetlink_dump_mark(a, b) (0)
 211 #endif
 212
 213 static inline int
 214 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 215 {
 216         u_int32_t id = htonl(ct->id);
 217         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 218         return 0;
 219
 220 nfattr_failure:
 221         return -1;
 222 }
 223
 224 static inline int
 225 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 226 {
 227         unsigned int use = htonl(atomic_read(&ct->ct_general.use));
 228
 229         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 230         return 0;
 231
 232 nfattr_failure:
 233         return -1;
 234 }
 235
 236 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 237
 238 static int
 239 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 240                     int event, int nowait,
 241                     const struct ip_conntrack *ct)
 242 {
 243         struct nlmsghdr *nlh;
 244         struct nfgenmsg *nfmsg;
 245         struct nfattr *nest_parms;
 246         unsigned char *b;
 247
 248         b = skb->tail;
 249
 250         event |= NFNL_SUBSYS_CTNETLINK << 8;
 251         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 252         nfmsg  = NLMSG_DATA(nlh);
 253
 254         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 255         nfmsg->nfgen_family = AF_INET;
 256         nfmsg->version      = NFNETLINK_V0;
 257         nfmsg->res_id       = 0;
 258
 259         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 260         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 261                 goto nfattr_failure;
 262         NFA_NEST_END(skb, nest_parms);
 263
 264         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 265         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 266                 goto nfattr_failure;
 267         NFA_NEST_END(skb, nest_parms);
 268
 269         if (ctnetlink_dump_status(skb, ct) < 0 ||
 270             ctnetlink_dump_timeout(skb, ct) < 0 ||
 271             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 272             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 273             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 274             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 275             ctnetlink_dump_mark(skb, ct) < 0 ||
 276             ctnetlink_dump_id(skb, ct) < 0 ||
 277             ctnetlink_dump_use(skb, ct) < 0)
 278                 goto nfattr_failure;
 279
 280         nlh->nlmsg_len = skb->tail - b;
 281         return skb->len;
 282
 283 nlmsg_failure:
 284 nfattr_failure:
 285         skb_trim(skb, b - skb->data);
 286         return -1;
 287 }
 288
 289 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 290 static int ctnetlink_conntrack_event(struct notifier_block *this,
 291                                      unsigned long events, void *ptr)
 292 {
 293         struct nlmsghdr *nlh;
 294         struct nfgenmsg *nfmsg;
 295         struct nfattr *nest_parms;
 296         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 297         struct sk_buff *skb;
 298         unsigned int type;
 299         unsigned char *b;
 300         unsigned int flags = 0, groups;
 301
 302         /* ignore our fake conntrack entry */
 303         if (ct == &ip_conntrack_untracked)
 304                 return NOTIFY_DONE;
 305
 306         if (events & IPCT_DESTROY) {
 307                 type = IPCTNL_MSG_CT_DELETE;
 308                 groups = NF_NETLINK_CONNTRACK_DESTROY;
 309                 goto alloc_skb;
 310         }
 311         if (events & (IPCT_NEW | IPCT_RELATED)) {
 312                 type = IPCTNL_MSG_CT_NEW;
 313                 flags = NLM_F_CREATE|NLM_F_EXCL;
 314                 /* dump everything */
 315                 events = ~0UL;
 316                 groups = NF_NETLINK_CONNTRACK_NEW;
 317                 goto alloc_skb;
 318         }
 319         if (events & (IPCT_STATUS |
 320                       IPCT_PROTOINFO |
 321                       IPCT_HELPER |
 322                       IPCT_HELPINFO |
 323                       IPCT_NATINFO)) {
 324                 type = IPCTNL_MSG_CT_NEW;
 325                 groups = NF_NETLINK_CONNTRACK_UPDATE;
 326                 goto alloc_skb;
 327         }
 328
 329         return NOTIFY_DONE;
 330
 331 alloc_skb:
 332   /* FIXME: Check if there are any listeners before, don't hurt performance */
 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         nlh->nlmsg_len = skb->tail - b;
 378         nfnetlink_send(skb, 0, groups, 0);
 379         return NOTIFY_DONE;
 380
 381 nlmsg_failure:
 382 nfattr_failure:
 383         kfree_skb(skb);
 384         return NOTIFY_DONE;
 385 }
 386 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 387
 388 static int ctnetlink_done(struct netlink_callback *cb)
 389 {
 390         DEBUGP("entered %s\n", __FUNCTION__);
 391         return 0;
 392 }
 393
 394 static int
 395 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 396 {
 397         struct ip_conntrack *ct = NULL;
 398         struct ip_conntrack_tuple_hash *h;
 399         struct list_head *i;
 400         u_int32_t *id = (u_int32_t *) &cb->args[1];
 401
 402         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 403                         cb->args[0], *id);
 404
 405         read_lock_bh(&ip_conntrack_lock);
 406         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 407                 list_for_each(i, &ip_conntrack_hash[cb->args[0]]) {
 408                         h = (struct ip_conntrack_tuple_hash *) i;
 409                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 410                                 continue;
 411                         ct = tuplehash_to_ctrack(h);
 412                         if (ct->id <= *id)
 413                                 continue;
 414                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 415                                                 cb->nlh->nlmsg_seq,
 416                                                 IPCTNL_MSG_CT_NEW,
 417                                                 1, ct) < 0)
 418                                 goto out;
 419                         *id = ct->id;
 420                 }
 421         }
 422 out:
 423         read_unlock_bh(&ip_conntrack_lock);
 424
 425         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 426
 427         return skb->len;
 428 }
 429
 430 #ifdef CONFIG_IP_NF_CT_ACCT
 431 static int
 432 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 433 {
 434         struct ip_conntrack *ct = NULL;
 435         struct ip_conntrack_tuple_hash *h;
 436         struct list_head *i;
 437         u_int32_t *id = (u_int32_t *) &cb->args[1];
 438
 439         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 440                         cb->args[0], *id);
 441
 442         write_lock_bh(&ip_conntrack_lock);
 443         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 444                 list_for_each(i, &ip_conntrack_hash[cb->args[0]]) {
 445                         h = (struct ip_conntrack_tuple_hash *) i;
 446                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 447                                 continue;
 448                         ct = tuplehash_to_ctrack(h);
 449                         if (ct->id <= *id)
 450                                 continue;
 451                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 452                                                 cb->nlh->nlmsg_seq,
 453                                                 IPCTNL_MSG_CT_NEW,
 454                                                 1, ct) < 0)
 455                                 goto out;
 456                         *id = ct->id;
 457
 458                         memset(&ct->counters, 0, sizeof(ct->counters));
 459                 }
 460         }
 461 out:
 462         write_unlock_bh(&ip_conntrack_lock);
 463
 464         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 465
 466         return skb->len;
 467 }
 468 #endif
 469
 470 static const int cta_min_ip[CTA_IP_MAX] = {
 471         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 472         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 473 };
 474
 475 static inline int
 476 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 477 {
 478         struct nfattr *tb[CTA_IP_MAX];
 479
 480         DEBUGP("entered %s\n", __FUNCTION__);
 481
 482         memset(tb, 0, CTA_IP_MAX * sizeof(tb));
 483
 484         if (nfattr_parse_nested(tb, CTA_IP_MAX, attr) < 0)
 485                 goto nfattr_failure;
 486
 487         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 488                 return -EINVAL;
 489
 490         if (!tb[CTA_IP_V4_SRC-1])
 491                 return -EINVAL;
 492         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 493
 494         if (!tb[CTA_IP_V4_DST-1])
 495                 return -EINVAL;
 496         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 497
 498         DEBUGP("leaving\n");
 499
 500         return 0;
 501
 502 nfattr_failure:
 503         return -1;
 504 }
 505
 506 static const int cta_min_proto[CTA_PROTO_MAX] = {
 507         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 508         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 509         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 510         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 511         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 512         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 513 };
 514
 515 static inline int
 516 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 517                             struct ip_conntrack_tuple *tuple)
 518 {
 519         struct nfattr *tb[CTA_PROTO_MAX];
 520         struct ip_conntrack_protocol *proto;
 521         int ret = 0;
 522
 523         DEBUGP("entered %s\n", __FUNCTION__);
 524
 525         memset(tb, 0, CTA_PROTO_MAX * sizeof(tb));
 526
 527         if (nfattr_parse_nested(tb, CTA_PROTO_MAX, attr) < 0)
 528                 goto nfattr_failure;
 529
 530         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 531                 return -EINVAL;
 532
 533         if (!tb[CTA_PROTO_NUM-1])
 534                 return -EINVAL;
 535         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 536
 537         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 538
 539         if (likely(proto && proto->nfattr_to_tuple)) {
 540                 ret = proto->nfattr_to_tuple(tb, tuple);
 541                 ip_conntrack_proto_put(proto);
 542         }
 543
 544         return ret;
 545
 546 nfattr_failure:
 547         return -1;
 548 }
 549
 550 static inline int
 551 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 552                       enum ctattr_tuple type)
 553 {
 554         struct nfattr *tb[CTA_TUPLE_MAX];
 555         int err;
 556
 557         DEBUGP("entered %s\n", __FUNCTION__);
 558
 559         memset(tb, 0, CTA_TUPLE_MAX * sizeof(tb));
 560         memset(tuple, 0, sizeof(*tuple));
 561
 562         if (nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]) < 0)
 563                 goto nfattr_failure;
 564
 565         if (!tb[CTA_TUPLE_IP-1])
 566                 return -EINVAL;
 567
 568         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 569         if (err < 0)
 570                 return err;
 571
 572         if (!tb[CTA_TUPLE_PROTO-1])
 573                 return -EINVAL;
 574
 575         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 576         if (err < 0)
 577                 return err;
 578
 579         /* orig and expect tuples get DIR_ORIGINAL */
 580         if (type == CTA_TUPLE_REPLY)
 581                 tuple->dst.dir = IP_CT_DIR_REPLY;
 582         else
 583                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 584
 585         DUMP_TUPLE(tuple);
 586
 587         DEBUGP("leaving\n");
 588
 589         return 0;
 590
 591 nfattr_failure:
 592         return -1;
 593 }
 594
 595 #ifdef CONFIG_IP_NF_NAT_NEEDED
 596 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
 597         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 598         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 599 };
 600
 601 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 602                                      const struct ip_conntrack *ct,
 603                                      struct ip_nat_range *range)
 604 {
 605         struct nfattr *tb[CTA_PROTONAT_MAX];
 606         struct ip_nat_protocol *npt;
 607
 608         DEBUGP("entered %s\n", __FUNCTION__);
 609
 610         memset(tb, 0, CTA_PROTONAT_MAX * sizeof(tb));
 611
 612         if (nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr) < 0)
 613                 goto nfattr_failure;
 614
 615         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 616                 goto nfattr_failure;
 617
 618         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 619         if (!npt)
 620                 return 0;
 621
 622         if (!npt->nfattr_to_range) {
 623                 ip_nat_proto_put(npt);
 624                 return 0;
 625         }
 626
 627         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 628         if (npt->nfattr_to_range(tb, range) > 0)
 629                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 630
 631         ip_nat_proto_put(npt);
 632
 633         DEBUGP("leaving\n");
 634         return 0;
 635
 636 nfattr_failure:
 637         return -1;
 638 }
 639
 640 static inline int
 641 ctnetlink_parse_nat(struct nfattr *cda[],
 642                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 643 {
 644         struct nfattr *tb[CTA_NAT_MAX];
 645         int err;
 646
 647         DEBUGP("entered %s\n", __FUNCTION__);
 648
 649         memset(tb, 0, CTA_NAT_MAX * sizeof(tb));
 650         memset(range, 0, sizeof(*range));
 651
 652         if (nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]) < 0)
 653                 goto nfattr_failure;
 654
 655         if (tb[CTA_NAT_MINIP-1])
 656                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 657
 658         if (!tb[CTA_NAT_MAXIP-1])
 659                 range->max_ip = range->min_ip;
 660         else
 661                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 662
 663         if (range->min_ip)
 664                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 665
 666         if (!tb[CTA_NAT_PROTO-1])
 667                 return 0;
 668
 669         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 670         if (err < 0)
 671                 return err;
 672
 673         DEBUGP("leaving\n");
 674         return 0;
 675
 676 nfattr_failure:
 677         return -1;
 678 }
 679 #endif
 680
 681 static inline int
 682 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 683 {
 684         struct nfattr *tb[CTA_HELP_MAX];
 685
 686         DEBUGP("entered %s\n", __FUNCTION__);
 687         memset(tb, 0, CTA_HELP_MAX * sizeof(tb));
 688
 689         if (nfattr_parse_nested(tb, CTA_HELP_MAX, attr) < 0)
 690                 goto nfattr_failure;
 691
 692         if (!tb[CTA_HELP_NAME-1])
 693                 return -EINVAL;
 694
 695         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 696
 697         return 0;
 698
 699 nfattr_failure:
 700         return -1;
 701 }
 702
 703 static int
 704 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 705                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 706 {
 707         struct ip_conntrack_tuple_hash *h;
 708         struct ip_conntrack_tuple tuple;
 709         struct ip_conntrack *ct;
 710         int err = 0;
 711
 712         DEBUGP("entered %s\n", __FUNCTION__);
 713
 714         if (cda[CTA_TUPLE_ORIG-1])
 715                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 716         else if (cda[CTA_TUPLE_REPLY-1])
 717                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 718         else {
 719                 /* Flush the whole table */
 720                 ip_conntrack_flush();
 721                 return 0;
 722         }
 723
 724         if (err < 0)
 725                 return err;
 726
 727         h = ip_conntrack_find_get(&tuple, NULL);
 728         if (!h) {
 729                 DEBUGP("tuple not found in conntrack hash\n");
 730                 return -ENOENT;
 731         }
 732
 733         ct = tuplehash_to_ctrack(h);
 734
 735         if (cda[CTA_ID-1]) {
 736                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 737                 if (ct->id != id) {
 738                         ip_conntrack_put(ct);
 739                         return -ENOENT;
 740                 }
 741         }
 742         if (del_timer(&ct->timeout)) {
 743                 ip_conntrack_put(ct);
 744                 ct->timeout.function((unsigned long)ct);
 745                 return 0;
 746         }
 747         ip_conntrack_put(ct);
 748         DEBUGP("leaving\n");
 749
 750         return 0;
 751 }
 752
 753 static int
 754 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 755                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 756 {
 757         struct ip_conntrack_tuple_hash *h;
 758         struct ip_conntrack_tuple tuple;
 759         struct ip_conntrack *ct;
 760         struct sk_buff *skb2 = NULL;
 761         int err = 0;
 762
 763         DEBUGP("entered %s\n", __FUNCTION__);
 764
 765         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 766                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 767                 u32 rlen;
 768
 769                 if (msg->nfgen_family != AF_INET)
 770                         return -EAFNOSUPPORT;
 771
 772                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 773                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 774 #ifdef CONFIG_IP_NF_CT_ACCT
 775                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 776                                                 ctnetlink_dump_table_w,
 777                                                 ctnetlink_done)) != 0)
 778                                 return -EINVAL;
 779 #else
 780                         return -ENOTSUPP;
 781 #endif
 782                 } else {
 783                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 784                                                         ctnetlink_dump_table,
 785                                                         ctnetlink_done)) != 0)
 786                         return -EINVAL;
 787                 }
 788
 789                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 790                 if (rlen > skb->len)
 791                         rlen = skb->len;
 792                 skb_pull(skb, rlen);
 793                 return 0;
 794         }
 795
 796         if (cda[CTA_TUPLE_ORIG-1])
 797                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 798         else if (cda[CTA_TUPLE_REPLY-1])
 799                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 800         else
 801                 return -EINVAL;
 802
 803         if (err < 0)
 804                 return err;
 805
 806         h = ip_conntrack_find_get(&tuple, NULL);
 807         if (!h) {
 808                 DEBUGP("tuple not found in conntrack hash");
 809                 return -ENOENT;
 810         }
 811         DEBUGP("tuple found\n");
 812         ct = tuplehash_to_ctrack(h);
 813
 814         err = -ENOMEM;
 815         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 816         if (!skb2) {
 817                 ip_conntrack_put(ct);
 818                 return -ENOMEM;
 819         }
 820         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 821
 822         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 823                                   IPCTNL_MSG_CT_NEW, 1, ct);
 824         ip_conntrack_put(ct);
 825         if (err <= 0)
 826                 goto out;
 827
 828         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 829         if (err < 0)
 830                 goto out;
 831
 832         DEBUGP("leaving\n");
 833         return 0;
 834
 835 out:
 836         if (skb2)
 837                 kfree_skb(skb2);
 838         return -1;
 839 }
 840
 841 static inline int
 842 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 843 {
 844         unsigned long d, status = *(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]);
 845         d = ct->status ^ status;
 846
 847         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 848                 /* unchangeable */
 849                 return -EINVAL;
 850
 851         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 852                 /* SEEN_REPLY bit can only be set */
 853                 return -EINVAL;
 854
 855
 856         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 857                 /* ASSURED bit can only be set */
 858                 return -EINVAL;
 859
 860         if (cda[CTA_NAT-1]) {
 861 #ifndef CONFIG_IP_NF_NAT_NEEDED
 862                 return -EINVAL;
 863 #else
 864                 unsigned int hooknum;
 865                 struct ip_nat_range range;
 866
 867                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 868                         return -EINVAL;
 869
 870                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 871                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 872                        htons(range.min.all), htons(range.max.all));
 873
 874                 /* This is tricky but it works. ip_nat_setup_info needs the
 875                  * hook number as parameter, so let's do the correct
 876                  * conversion and run away */
 877                 if (status & IPS_SRC_NAT_DONE)
 878                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 879                 else if (status & IPS_DST_NAT_DONE)
 880                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 881                 else
 882                         return -EINVAL; /* Missing NAT flags */
 883
 884                 DEBUGP("NAT status: %lu\n",
 885                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 886
 887                 if (ip_nat_initialized(ct, hooknum))
 888                         return -EEXIST;
 889                 ip_nat_setup_info(ct, &range, hooknum);
 890
 891                 DEBUGP("NAT status after setup_info: %lu\n",
 892                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 893 #endif
 894         }
 895
 896         /* Be careful here, modifying NAT bits can screw up things,
 897          * so don't let users modify them directly if they don't pass
 898          * ip_nat_range. */
 899         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 900         return 0;
 901 }
 902
 903
 904 static inline int
 905 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 906 {
 907         struct ip_conntrack_helper *helper;
 908         char *helpname;
 909         int err;
 910
 911         DEBUGP("entered %s\n", __FUNCTION__);
 912
 913         /* don't change helper of sibling connections */
 914         if (ct->master)
 915                 return -EINVAL;
 916
 917         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 918         if (err < 0)
 919                 return err;
 920
 921         helper = __ip_conntrack_helper_find_byname(helpname);
 922         if (!helper) {
 923                 if (!strcmp(helpname, ""))
 924                         helper = NULL;
 925                 else
 926                         return -EINVAL;
 927         }
 928
 929         if (ct->helper) {
 930                 if (!helper) {
 931                         /* we had a helper before ... */
 932                         ip_ct_remove_expectations(ct);
 933                         ct->helper = NULL;
 934                 } else {
 935                         /* need to zero data of old helper */
 936                         memset(&ct->help, 0, sizeof(ct->help));
 937                 }
 938         }
 939
 940         ct->helper = helper;
 941
 942         return 0;
 943 }
 944
 945 static inline int
 946 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 947 {
 948         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 949
 950         if (!del_timer(&ct->timeout))
 951                 return -ETIME;
 952
 953         ct->timeout.expires = jiffies + timeout * HZ;
 954         add_timer(&ct->timeout);
 955
 956         return 0;
 957 }
 958
 959 static int
 960 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 961 {
 962         int err;
 963
 964         DEBUGP("entered %s\n", __FUNCTION__);
 965
 966         if (cda[CTA_HELP-1]) {
 967                 err = ctnetlink_change_helper(ct, cda);
 968                 if (err < 0)
 969                         return err;
 970         }
 971
 972         if (cda[CTA_TIMEOUT-1]) {
 973                 err = ctnetlink_change_timeout(ct, cda);
 974                 if (err < 0)
 975                         return err;
 976         }
 977
 978         if (cda[CTA_STATUS-1]) {
 979                 err = ctnetlink_change_status(ct, cda);
 980                 if (err < 0)
 981                         return err;
 982         }
 983
 984         DEBUGP("all done\n");
 985         return 0;
 986 }
 987
 988 static int
 989 ctnetlink_create_conntrack(struct nfattr *cda[],
 990                            struct ip_conntrack_tuple *otuple,
 991                            struct ip_conntrack_tuple *rtuple)
 992 {
 993         struct ip_conntrack *ct;
 994         int err = -EINVAL;
 995
 996         DEBUGP("entered %s\n", __FUNCTION__);
 997
 998         ct = ip_conntrack_alloc(otuple, rtuple);
 999         if (ct == NULL || IS_ERR(ct))
1000                 return -ENOMEM;
1001
1002         if (!cda[CTA_TIMEOUT-1])
1003                 goto err;
1004         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1005
1006         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1007         ct->status |= IPS_CONFIRMED;
1008
1009         err = ctnetlink_change_status(ct, cda);
1010         if (err < 0)
1011                 goto err;
1012
1013         ct->helper = ip_conntrack_helper_find_get(rtuple);
1014
1015         add_timer(&ct->timeout);
1016         ip_conntrack_hash_insert(ct);
1017
1018         if (ct->helper)
1019                 ip_conntrack_helper_put(ct->helper);
1020
1021         DEBUGP("conntrack with id %u inserted\n", ct->id);
1022         return 0;
1023
1024 err:
1025         ip_conntrack_free(ct);
1026         return err;
1027 }
1028
1029 static int
1030 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1031                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1032 {
1033         struct ip_conntrack_tuple otuple, rtuple;
1034         struct ip_conntrack_tuple_hash *h = NULL;
1035         int err = 0;
1036
1037         DEBUGP("entered %s\n", __FUNCTION__);
1038
1039         if (cda[CTA_TUPLE_ORIG-1]) {
1040                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1041                 if (err < 0)
1042                         return err;
1043         }
1044
1045         if (cda[CTA_TUPLE_REPLY-1]) {
1046                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1047                 if (err < 0)
1048                         return err;
1049         }
1050
1051         write_lock_bh(&ip_conntrack_lock);
1052         if (cda[CTA_TUPLE_ORIG-1])
1053                 h = __ip_conntrack_find(&otuple, NULL);
1054         else if (cda[CTA_TUPLE_REPLY-1])
1055                 h = __ip_conntrack_find(&rtuple, NULL);
1056
1057         if (h == NULL) {
1058                 write_unlock_bh(&ip_conntrack_lock);
1059                 DEBUGP("no such conntrack, create new\n");
1060                 err = -ENOENT;
1061                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1062                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1063                 goto out_unlock;
1064         } else {
1065                 /* we only allow nat config for new conntracks */
1066                 if (cda[CTA_NAT-1]) {
1067                         err = -EINVAL;
1068                         goto out_unlock;
1069                 }
1070         }
1071
1072         /* We manipulate the conntrack inside the global conntrack table lock,
1073          * so there's no need to increase the refcount */
1074         DEBUGP("conntrack found\n");
1075         err = -EEXIST;
1076         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1077                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1078
1079 out_unlock:
1080         write_unlock_bh(&ip_conntrack_lock);
1081         return err;
1082 }
1083
1084 /***********************************************************************
1085  * EXPECT
1086  ***********************************************************************/
1087
1088 static inline int
1089 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1090                          const struct ip_conntrack_tuple *tuple,
1091                          enum ctattr_expect type)
1092 {
1093         struct nfattr *nest_parms = NFA_NEST(skb, type);
1094
1095         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1096                 goto nfattr_failure;
1097
1098         NFA_NEST_END(skb, nest_parms);
1099
1100         return 0;
1101
1102 nfattr_failure:
1103         return -1;
1104 }
1105
1106 static inline int
1107 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1108                           const struct ip_conntrack_expect *exp)
1109 {
1110         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1111         u_int32_t id = htonl(exp->id);
1112         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT);
1113
1114         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1115                 goto nfattr_failure;
1116         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1117                 goto nfattr_failure;
1118
1119         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1120         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1121         NFA_NEST_END(skb, nest_parms);
1122
1123         return 0;
1124
1125 nfattr_failure:
1126         return -1;
1127 }
1128
1129 static int
1130 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1131                     int event,
1132                     int nowait,
1133                     const struct ip_conntrack_expect *exp)
1134 {
1135         struct nlmsghdr *nlh;
1136         struct nfgenmsg *nfmsg;
1137         unsigned char *b;
1138
1139         b = skb->tail;
1140
1141         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1142         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1143         nfmsg  = NLMSG_DATA(nlh);
1144
1145         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1146         nfmsg->nfgen_family = AF_INET;
1147         nfmsg->version      = NFNETLINK_V0;
1148         nfmsg->res_id       = 0;
1149
1150         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1151                 goto nfattr_failure;
1152
1153         nlh->nlmsg_len = skb->tail - b;
1154         return skb->len;
1155
1156 nlmsg_failure:
1157 nfattr_failure:
1158         skb_trim(skb, b - skb->data);
1159         return -1;
1160 }
1161
1162 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1163 static int ctnetlink_expect_event(struct notifier_block *this,
1164                                   unsigned long events, void *ptr)
1165 {
1166         struct nlmsghdr *nlh;
1167         struct nfgenmsg *nfmsg;
1168         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1169         struct sk_buff *skb;
1170         unsigned int type;
1171         unsigned char *b;
1172         int flags = 0;
1173         u16 proto;
1174
1175         if (events & IPEXP_NEW) {
1176                 type = IPCTNL_MSG_EXP_NEW;
1177                 flags = NLM_F_CREATE|NLM_F_EXCL;
1178         } else
1179                 return NOTIFY_DONE;
1180
1181         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1182         if (!skb)
1183                 return NOTIFY_DONE;
1184
1185         b = skb->tail;
1186
1187         type |= NFNL_SUBSYS_CTNETLINK << 8;
1188         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1189         nfmsg = NLMSG_DATA(nlh);
1190
1191         nlh->nlmsg_flags    = flags;
1192         nfmsg->nfgen_family = AF_INET;
1193         nfmsg->version      = NFNETLINK_V0;
1194         nfmsg->res_id       = 0;
1195
1196         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1197                 goto nfattr_failure;
1198
1199         nlh->nlmsg_len = skb->tail - b;
1200         proto = exp->tuple.dst.protonum;
1201         nfnetlink_send(skb, 0, NF_NETLINK_CONNTRACK_EXP_NEW, 0);
1202         return NOTIFY_DONE;
1203
1204 nlmsg_failure:
1205 nfattr_failure:
1206         kfree_skb(skb);
1207         return NOTIFY_DONE;
1208 }
1209 #endif
1210
1211 static int
1212 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1213 {
1214         struct ip_conntrack_expect *exp = NULL;
1215         struct list_head *i;
1216         u_int32_t *id = (u_int32_t *) &cb->args[0];
1217
1218         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1219
1220         read_lock_bh(&ip_conntrack_lock);
1221         list_for_each(i, &ip_conntrack_expect_list) {
1222                 exp = (struct ip_conntrack_expect *) i;
1223                 if (exp->id <= *id)
1224                         continue;
1225                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1226                                             cb->nlh->nlmsg_seq,
1227                                             IPCTNL_MSG_EXP_NEW,
1228                                             1, exp) < 0)
1229                         goto out;
1230                 *id = exp->id;
1231         }
1232 out:
1233         read_unlock_bh(&ip_conntrack_lock);
1234
1235         DEBUGP("leaving, last id=%llu\n", *id);
1236
1237         return skb->len;
1238 }
1239
1240 static int
1241 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1242                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1243 {
1244         struct ip_conntrack_tuple tuple;
1245         struct ip_conntrack_expect *exp;
1246         struct sk_buff *skb2;
1247         int err = 0;
1248
1249         DEBUGP("entered %s\n", __FUNCTION__);
1250
1251         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1252                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1253                 u32 rlen;
1254
1255                 if (msg->nfgen_family != AF_INET)
1256                         return -EAFNOSUPPORT;
1257
1258                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1259                                                 ctnetlink_exp_dump_table,
1260                                                 ctnetlink_done)) != 0)
1261                         return -EINVAL;
1262                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1263                 if (rlen > skb->len)
1264                         rlen = skb->len;
1265                 skb_pull(skb, rlen);
1266                 return 0;
1267         }
1268
1269         if (cda[CTA_TUPLE_ORIG-1])
1270                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
1271         else if (cda[CTA_TUPLE_REPLY-1])
1272                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
1273         else
1274                 return -EINVAL;
1275
1276         if (err < 0)
1277                 return err;
1278
1279         exp = ip_conntrack_expect_find_get(&tuple);
1280         if (!exp)
1281                 return -ENOENT;
1282
1283         err = -ENOMEM;
1284         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1285         if (!skb2)
1286                 goto out;
1287         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1288
1289         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1290                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1291                                       1, exp);
1292         if (err <= 0)
1293                 goto out;
1294
1295         ip_conntrack_expect_put(exp);
1296
1297         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1298         if (err < 0)
1299                 goto free;
1300
1301         return err;
1302
1303 out:
1304         ip_conntrack_expect_put(exp);
1305 free:
1306         if (skb2)
1307                 kfree_skb(skb2);
1308         return err;
1309 }
1310
1311 static int
1312 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1313                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1314 {
1315         struct ip_conntrack_expect *exp, *tmp;
1316         struct ip_conntrack_tuple tuple;
1317         struct ip_conntrack_helper *h;
1318         int err;
1319
1320         /* delete by tuple needs either orig or reply tuple */
1321         if (cda[CTA_TUPLE_ORIG-1])
1322                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
1323         else if (cda[CTA_TUPLE_REPLY-1])
1324                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
1325         else if (cda[CTA_HELP_NAME-1]) {
1326                 char *name = NFA_DATA(cda[CTA_HELP_NAME-1]);
1327
1328                 /* delete all expectations for this helper */
1329                 write_lock_bh(&ip_conntrack_lock);
1330                 h = __ip_conntrack_helper_find_byname(name);
1331                 if (!h) {
1332                         write_unlock_bh(&ip_conntrack_lock);
1333                         return -EINVAL;
1334                 }
1335                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1336                                          list) {
1337                         if (exp->master->helper == h
1338                             && del_timer(&exp->timeout))
1339                                 __ip_ct_expect_unlink_destroy(exp);
1340                 }
1341                 write_unlock(&ip_conntrack_lock);
1342                 return 0;
1343         } else {
1344                 /* This basically means we have to flush everything*/
1345                 write_lock_bh(&ip_conntrack_lock);
1346                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1347                                          list) {
1348                         if (del_timer(&exp->timeout))
1349                                 __ip_ct_expect_unlink_destroy(exp);
1350                 }
1351                 write_unlock_bh(&ip_conntrack_lock);
1352                 return 0;
1353         }
1354
1355         if (err < 0)
1356                 return err;
1357
1358         /* bump usage count to 2 */
1359         exp = ip_conntrack_expect_find_get(&tuple);
1360         if (!exp)
1361                 return -ENOENT;
1362
1363         if (cda[CTA_EXPECT_ID-1]) {
1364                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1365                 if (exp->id != ntohl(id)) {
1366                         ip_conntrack_expect_put(exp);
1367                         return -ENOENT;
1368                 }
1369         }
1370
1371         /* after list removal, usage count == 1 */
1372         ip_conntrack_unexpect_related(exp);
1373         /* have to put what we 'get' above. after this line usage count == 0 */
1374         ip_conntrack_expect_put(exp);
1375
1376         return 0;
1377 }
1378 static int
1379 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1380 {
1381         return -EOPNOTSUPP;
1382 }
1383
1384 static int
1385 ctnetlink_create_expect(struct nfattr *cda[])
1386 {
1387         struct ip_conntrack_tuple tuple, mask, master_tuple;
1388         struct ip_conntrack_tuple_hash *h = NULL;
1389         struct ip_conntrack_expect *exp;
1390         struct ip_conntrack *ct;
1391         int err = 0;
1392
1393         DEBUGP("entered %s\n", __FUNCTION__);
1394
1395         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1396         if (err < 0)
1397                 return err;
1398         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASK);
1399         if (err < 0)
1400                 return err;
1401
1402         if (cda[CTA_TUPLE_ORIG-1])
1403                 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_TUPLE_ORIG);
1404         else if (cda[CTA_TUPLE_REPLY-1])
1405                 err = ctnetlink_parse_tuple(cda, &master_tuple,
1406                                             CTA_TUPLE_REPLY);
1407         else
1408                 return -EINVAL;
1409
1410         if (err < 0)
1411                 return err;
1412
1413         /* Look for master conntrack of this expectation */
1414         h = ip_conntrack_find_get(&master_tuple, NULL);
1415         if (!h)
1416                 return -ENOENT;
1417         ct = tuplehash_to_ctrack(h);
1418
1419         if (!ct->helper) {
1420                 /* such conntrack hasn't got any helper, abort */
1421                 err = -EINVAL;
1422                 goto out;
1423         }
1424
1425         exp = ip_conntrack_expect_alloc(ct);
1426         if (!exp) {
1427                 err = -ENOMEM;
1428                 goto out;
1429         }
1430
1431         exp->expectfn = NULL;
1432         exp->master = ct;
1433         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1434         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1435
1436         err = ip_conntrack_expect_related(exp);
1437         ip_conntrack_expect_put(exp);
1438
1439 out:
1440         ip_conntrack_put(tuplehash_to_ctrack(h));
1441         return err;
1442 }
1443
1444 static int
1445 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1446                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1447 {
1448         struct ip_conntrack_tuple tuple;
1449         struct ip_conntrack_expect *exp;
1450         int err = 0;
1451
1452         DEBUGP("entered %s\n", __FUNCTION__);
1453
1454         if (!cda[CTA_EXPECT_TUPLE-1] || !cda[CTA_EXPECT_MASK-1])
1455                 return -EINVAL;
1456
1457         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1458         if (err < 0)
1459                 return err;
1460
1461         write_lock_bh(&ip_conntrack_lock);
1462         exp = __ip_conntrack_expect_find(&tuple);
1463
1464         if (!exp) {
1465                 write_unlock_bh(&ip_conntrack_lock);
1466                 err = -ENOENT;
1467                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1468                         err = ctnetlink_create_expect(cda);
1469                 return err;
1470         }
1471
1472         err = -EEXIST;
1473         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1474                 err = ctnetlink_change_expect(exp, cda);
1475         write_unlock_bh(&ip_conntrack_lock);
1476
1477         DEBUGP("leaving\n");
1478
1479         return err;
1480 }
1481
1482 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1483 static struct notifier_block ctnl_notifier = {
1484         .notifier_call  = ctnetlink_conntrack_event,
1485 };
1486
1487 static struct notifier_block ctnl_notifier_exp = {
1488         .notifier_call  = ctnetlink_expect_event,
1489 };
1490 #endif
1491
1492 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1493         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1494                                             .cap_required = CAP_NET_ADMIN },
1495         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1496                                             .cap_required = CAP_NET_ADMIN },
1497         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1498                                             .cap_required = CAP_NET_ADMIN },
1499         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1500                                             .cap_required = CAP_NET_ADMIN },
1501 };
1502
1503 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_MAX] = {
1504         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1505                                             .cap_required = CAP_NET_ADMIN },
1506         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1507                                             .cap_required = CAP_NET_ADMIN },
1508         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1509                                             .cap_required = CAP_NET_ADMIN },
1510 };
1511
1512 static struct nfnetlink_subsystem ctnl_subsys = {
1513         .name                           = "conntrack",
1514         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1515         .cb_count                       = IPCTNL_MSG_MAX,
1516         .attr_count                     = CTA_MAX,
1517         .cb                             = ctnl_cb,
1518 };
1519
1520 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1521         .name                           = "conntrack_expect",
1522         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1523         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1524         .attr_count                     = CTA_MAX,
1525         .cb                             = ctnl_exp_cb,
1526 };
1527
1528 static int __init ctnetlink_init(void)
1529 {
1530         int ret;
1531
1532         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1533         ret = nfnetlink_subsys_register(&ctnl_subsys);
1534         if (ret < 0) {
1535                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1536                 goto err_out;
1537         }
1538
1539         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1540         if (ret < 0) {
1541                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1542                 goto err_unreg_subsys;
1543         }
1544
1545 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1546         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1547         if (ret < 0) {
1548                 printk("ctnetlink_init: cannot register notifier.\n");
1549                 goto err_unreg_exp_subsys;
1550         }
1551
1552         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1553         if (ret < 0) {
1554                 printk("ctnetlink_init: cannot expect register notifier.\n");
1555                 goto err_unreg_notifier;
1556         }
1557 #endif
1558
1559         return 0;
1560
1561 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1562 err_unreg_notifier:
1563         ip_conntrack_unregister_notifier(&ctnl_notifier);
1564 err_unreg_exp_subsys:
1565         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1566 #endif
1567 err_unreg_subsys:
1568         nfnetlink_subsys_unregister(&ctnl_subsys);
1569 err_out:
1570         return ret;
1571 }
1572
1573 static void __exit ctnetlink_exit(void)
1574 {
1575         printk("ctnetlink: unregistering from nfnetlink.\n");
1576
1577 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1578         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1579         ip_conntrack_unregister_notifier(&ctnl_notifier);
1580 #endif
1581
1582         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1583         nfnetlink_subsys_unregister(&ctnl_subsys);
1584         return;
1585 }
1586
1587 module_init(ctnetlink_init);
1588 module_exit(ctnetlink_exit);