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, group;
 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                 group = NFNLGRP_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                 group = NFNLGRP_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                 group = NFNLGRP_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, group, 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_prev(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_prev(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
 483         if (nfattr_parse_nested(tb, CTA_IP_MAX, attr) < 0)
 484                 goto nfattr_failure;
 485
 486         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 487                 return -EINVAL;
 488
 489         if (!tb[CTA_IP_V4_SRC-1])
 490                 return -EINVAL;
 491         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 492
 493         if (!tb[CTA_IP_V4_DST-1])
 494                 return -EINVAL;
 495         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 496
 497         DEBUGP("leaving\n");
 498
 499         return 0;
 500
 501 nfattr_failure:
 502         return -1;
 503 }
 504
 505 static const int cta_min_proto[CTA_PROTO_MAX] = {
 506         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 507         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 508         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 509         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 510         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 511         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 512 };
 513
 514 static inline int
 515 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 516                             struct ip_conntrack_tuple *tuple)
 517 {
 518         struct nfattr *tb[CTA_PROTO_MAX];
 519         struct ip_conntrack_protocol *proto;
 520         int ret = 0;
 521
 522         DEBUGP("entered %s\n", __FUNCTION__);
 523
 524         if (nfattr_parse_nested(tb, CTA_PROTO_MAX, attr) < 0)
 525                 goto nfattr_failure;
 526
 527         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 528                 return -EINVAL;
 529
 530         if (!tb[CTA_PROTO_NUM-1])
 531                 return -EINVAL;
 532         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 533
 534         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 535
 536         if (likely(proto && proto->nfattr_to_tuple)) {
 537                 ret = proto->nfattr_to_tuple(tb, tuple);
 538                 ip_conntrack_proto_put(proto);
 539         }
 540
 541         return ret;
 542
 543 nfattr_failure:
 544         return -1;
 545 }
 546
 547 static inline int
 548 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 549                       enum ctattr_tuple type)
 550 {
 551         struct nfattr *tb[CTA_TUPLE_MAX];
 552         int err;
 553
 554         DEBUGP("entered %s\n", __FUNCTION__);
 555
 556         memset(tuple, 0, sizeof(*tuple));
 557
 558         if (nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]) < 0)
 559                 goto nfattr_failure;
 560
 561         if (!tb[CTA_TUPLE_IP-1])
 562                 return -EINVAL;
 563
 564         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 565         if (err < 0)
 566                 return err;
 567
 568         if (!tb[CTA_TUPLE_PROTO-1])
 569                 return -EINVAL;
 570
 571         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 572         if (err < 0)
 573                 return err;
 574
 575         /* orig and expect tuples get DIR_ORIGINAL */
 576         if (type == CTA_TUPLE_REPLY)
 577                 tuple->dst.dir = IP_CT_DIR_REPLY;
 578         else
 579                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 580
 581         DUMP_TUPLE(tuple);
 582
 583         DEBUGP("leaving\n");
 584
 585         return 0;
 586
 587 nfattr_failure:
 588         return -1;
 589 }
 590
 591 #ifdef CONFIG_IP_NF_NAT_NEEDED
 592 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
 593         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 594         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 595 };
 596
 597 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 598                                      const struct ip_conntrack *ct,
 599                                      struct ip_nat_range *range)
 600 {
 601         struct nfattr *tb[CTA_PROTONAT_MAX];
 602         struct ip_nat_protocol *npt;
 603
 604         DEBUGP("entered %s\n", __FUNCTION__);
 605
 606         if (nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr) < 0)
 607                 goto nfattr_failure;
 608
 609         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 610                 goto nfattr_failure;
 611
 612         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 613         if (!npt)
 614                 return 0;
 615
 616         if (!npt->nfattr_to_range) {
 617                 ip_nat_proto_put(npt);
 618                 return 0;
 619         }
 620
 621         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 622         if (npt->nfattr_to_range(tb, range) > 0)
 623                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 624
 625         ip_nat_proto_put(npt);
 626
 627         DEBUGP("leaving\n");
 628         return 0;
 629
 630 nfattr_failure:
 631         return -1;
 632 }
 633
 634 static inline int
 635 ctnetlink_parse_nat(struct nfattr *cda[],
 636                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 637 {
 638         struct nfattr *tb[CTA_NAT_MAX];
 639         int err;
 640
 641         DEBUGP("entered %s\n", __FUNCTION__);
 642
 643         memset(range, 0, sizeof(*range));
 644
 645         if (nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]) < 0)
 646                 goto nfattr_failure;
 647
 648         if (tb[CTA_NAT_MINIP-1])
 649                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 650
 651         if (!tb[CTA_NAT_MAXIP-1])
 652                 range->max_ip = range->min_ip;
 653         else
 654                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 655
 656         if (range->min_ip)
 657                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 658
 659         if (!tb[CTA_NAT_PROTO-1])
 660                 return 0;
 661
 662         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 663         if (err < 0)
 664                 return err;
 665
 666         DEBUGP("leaving\n");
 667         return 0;
 668
 669 nfattr_failure:
 670         return -1;
 671 }
 672 #endif
 673
 674 static inline int
 675 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 676 {
 677         struct nfattr *tb[CTA_HELP_MAX];
 678
 679         DEBUGP("entered %s\n", __FUNCTION__);
 680
 681         if (nfattr_parse_nested(tb, CTA_HELP_MAX, attr) < 0)
 682                 goto nfattr_failure;
 683
 684         if (!tb[CTA_HELP_NAME-1])
 685                 return -EINVAL;
 686
 687         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 688
 689         return 0;
 690
 691 nfattr_failure:
 692         return -1;
 693 }
 694
 695 static int
 696 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 697                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 698 {
 699         struct ip_conntrack_tuple_hash *h;
 700         struct ip_conntrack_tuple tuple;
 701         struct ip_conntrack *ct;
 702         int err = 0;
 703
 704         DEBUGP("entered %s\n", __FUNCTION__);
 705
 706         if (cda[CTA_TUPLE_ORIG-1])
 707                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 708         else if (cda[CTA_TUPLE_REPLY-1])
 709                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 710         else {
 711                 /* Flush the whole table */
 712                 ip_conntrack_flush();
 713                 return 0;
 714         }
 715
 716         if (err < 0)
 717                 return err;
 718
 719         h = ip_conntrack_find_get(&tuple, NULL);
 720         if (!h) {
 721                 DEBUGP("tuple not found in conntrack hash\n");
 722                 return -ENOENT;
 723         }
 724
 725         ct = tuplehash_to_ctrack(h);
 726
 727         if (cda[CTA_ID-1]) {
 728                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 729                 if (ct->id != id) {
 730                         ip_conntrack_put(ct);
 731                         return -ENOENT;
 732                 }
 733         }
 734         if (del_timer(&ct->timeout)) {
 735                 ip_conntrack_put(ct);
 736                 ct->timeout.function((unsigned long)ct);
 737                 return 0;
 738         }
 739         ip_conntrack_put(ct);
 740         DEBUGP("leaving\n");
 741
 742         return 0;
 743 }
 744
 745 static int
 746 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 747                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 748 {
 749         struct ip_conntrack_tuple_hash *h;
 750         struct ip_conntrack_tuple tuple;
 751         struct ip_conntrack *ct;
 752         struct sk_buff *skb2 = NULL;
 753         int err = 0;
 754
 755         DEBUGP("entered %s\n", __FUNCTION__);
 756
 757         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 758                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 759                 u32 rlen;
 760
 761                 if (msg->nfgen_family != AF_INET)
 762                         return -EAFNOSUPPORT;
 763
 764                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 765                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 766 #ifdef CONFIG_IP_NF_CT_ACCT
 767                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 768                                                 ctnetlink_dump_table_w,
 769                                                 ctnetlink_done)) != 0)
 770                                 return -EINVAL;
 771 #else
 772                         return -ENOTSUPP;
 773 #endif
 774                 } else {
 775                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 776                                                         ctnetlink_dump_table,
 777                                                         ctnetlink_done)) != 0)
 778                         return -EINVAL;
 779                 }
 780
 781                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 782                 if (rlen > skb->len)
 783                         rlen = skb->len;
 784                 skb_pull(skb, rlen);
 785                 return 0;
 786         }
 787
 788         if (cda[CTA_TUPLE_ORIG-1])
 789                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 790         else if (cda[CTA_TUPLE_REPLY-1])
 791                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 792         else
 793                 return -EINVAL;
 794
 795         if (err < 0)
 796                 return err;
 797
 798         h = ip_conntrack_find_get(&tuple, NULL);
 799         if (!h) {
 800                 DEBUGP("tuple not found in conntrack hash");
 801                 return -ENOENT;
 802         }
 803         DEBUGP("tuple found\n");
 804         ct = tuplehash_to_ctrack(h);
 805
 806         err = -ENOMEM;
 807         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 808         if (!skb2) {
 809                 ip_conntrack_put(ct);
 810                 return -ENOMEM;
 811         }
 812         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 813
 814         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 815                                   IPCTNL_MSG_CT_NEW, 1, ct);
 816         ip_conntrack_put(ct);
 817         if (err <= 0)
 818                 goto out;
 819
 820         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 821         if (err < 0)
 822                 goto out;
 823
 824         DEBUGP("leaving\n");
 825         return 0;
 826
 827 out:
 828         if (skb2)
 829                 kfree_skb(skb2);
 830         return -1;
 831 }
 832
 833 static inline int
 834 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 835 {
 836         unsigned long d, status = *(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]);
 837         d = ct->status ^ status;
 838
 839         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 840                 /* unchangeable */
 841                 return -EINVAL;
 842
 843         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 844                 /* SEEN_REPLY bit can only be set */
 845                 return -EINVAL;
 846
 847
 848         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 849                 /* ASSURED bit can only be set */
 850                 return -EINVAL;
 851
 852         if (cda[CTA_NAT-1]) {
 853 #ifndef CONFIG_IP_NF_NAT_NEEDED
 854                 return -EINVAL;
 855 #else
 856                 unsigned int hooknum;
 857                 struct ip_nat_range range;
 858
 859                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 860                         return -EINVAL;
 861
 862                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 863                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 864                        htons(range.min.all), htons(range.max.all));
 865
 866                 /* This is tricky but it works. ip_nat_setup_info needs the
 867                  * hook number as parameter, so let's do the correct
 868                  * conversion and run away */
 869                 if (status & IPS_SRC_NAT_DONE)
 870                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 871                 else if (status & IPS_DST_NAT_DONE)
 872                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 873                 else
 874                         return -EINVAL; /* Missing NAT flags */
 875
 876                 DEBUGP("NAT status: %lu\n",
 877                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 878
 879                 if (ip_nat_initialized(ct, hooknum))
 880                         return -EEXIST;
 881                 ip_nat_setup_info(ct, &range, hooknum);
 882
 883                 DEBUGP("NAT status after setup_info: %lu\n",
 884                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 885 #endif
 886         }
 887
 888         /* Be careful here, modifying NAT bits can screw up things,
 889          * so don't let users modify them directly if they don't pass
 890          * ip_nat_range. */
 891         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 892         return 0;
 893 }
 894
 895
 896 static inline int
 897 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 898 {
 899         struct ip_conntrack_helper *helper;
 900         char *helpname;
 901         int err;
 902
 903         DEBUGP("entered %s\n", __FUNCTION__);
 904
 905         /* don't change helper of sibling connections */
 906         if (ct->master)
 907                 return -EINVAL;
 908
 909         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 910         if (err < 0)
 911                 return err;
 912
 913         helper = __ip_conntrack_helper_find_byname(helpname);
 914         if (!helper) {
 915                 if (!strcmp(helpname, ""))
 916                         helper = NULL;
 917                 else
 918                         return -EINVAL;
 919         }
 920
 921         if (ct->helper) {
 922                 if (!helper) {
 923                         /* we had a helper before ... */
 924                         ip_ct_remove_expectations(ct);
 925                         ct->helper = NULL;
 926                 } else {
 927                         /* need to zero data of old helper */
 928                         memset(&ct->help, 0, sizeof(ct->help));
 929                 }
 930         }
 931
 932         ct->helper = helper;
 933
 934         return 0;
 935 }
 936
 937 static inline int
 938 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 939 {
 940         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 941
 942         if (!del_timer(&ct->timeout))
 943                 return -ETIME;
 944
 945         ct->timeout.expires = jiffies + timeout * HZ;
 946         add_timer(&ct->timeout);
 947
 948         return 0;
 949 }
 950
 951 static int
 952 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 953 {
 954         int err;
 955
 956         DEBUGP("entered %s\n", __FUNCTION__);
 957
 958         if (cda[CTA_HELP-1]) {
 959                 err = ctnetlink_change_helper(ct, cda);
 960                 if (err < 0)
 961                         return err;
 962         }
 963
 964         if (cda[CTA_TIMEOUT-1]) {
 965                 err = ctnetlink_change_timeout(ct, cda);
 966                 if (err < 0)
 967                         return err;
 968         }
 969
 970         if (cda[CTA_STATUS-1]) {
 971                 err = ctnetlink_change_status(ct, cda);
 972                 if (err < 0)
 973                         return err;
 974         }
 975
 976         DEBUGP("all done\n");
 977         return 0;
 978 }
 979
 980 static int
 981 ctnetlink_create_conntrack(struct nfattr *cda[],
 982                            struct ip_conntrack_tuple *otuple,
 983                            struct ip_conntrack_tuple *rtuple)
 984 {
 985         struct ip_conntrack *ct;
 986         int err = -EINVAL;
 987
 988         DEBUGP("entered %s\n", __FUNCTION__);
 989
 990         ct = ip_conntrack_alloc(otuple, rtuple);
 991         if (ct == NULL || IS_ERR(ct))
 992                 return -ENOMEM;
 993
 994         if (!cda[CTA_TIMEOUT-1])
 995                 goto err;
 996         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 997
 998         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
 999         ct->status |= IPS_CONFIRMED;
1000
1001         err = ctnetlink_change_status(ct, cda);
1002         if (err < 0)
1003                 goto err;
1004
1005         ct->helper = ip_conntrack_helper_find_get(rtuple);
1006
1007         add_timer(&ct->timeout);
1008         ip_conntrack_hash_insert(ct);
1009
1010         if (ct->helper)
1011                 ip_conntrack_helper_put(ct->helper);
1012
1013         DEBUGP("conntrack with id %u inserted\n", ct->id);
1014         return 0;
1015
1016 err:
1017         ip_conntrack_free(ct);
1018         return err;
1019 }
1020
1021 static int
1022 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1023                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1024 {
1025         struct ip_conntrack_tuple otuple, rtuple;
1026         struct ip_conntrack_tuple_hash *h = NULL;
1027         int err = 0;
1028
1029         DEBUGP("entered %s\n", __FUNCTION__);
1030
1031         if (cda[CTA_TUPLE_ORIG-1]) {
1032                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1033                 if (err < 0)
1034                         return err;
1035         }
1036
1037         if (cda[CTA_TUPLE_REPLY-1]) {
1038                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1039                 if (err < 0)
1040                         return err;
1041         }
1042
1043         write_lock_bh(&ip_conntrack_lock);
1044         if (cda[CTA_TUPLE_ORIG-1])
1045                 h = __ip_conntrack_find(&otuple, NULL);
1046         else if (cda[CTA_TUPLE_REPLY-1])
1047                 h = __ip_conntrack_find(&rtuple, NULL);
1048
1049         if (h == NULL) {
1050                 write_unlock_bh(&ip_conntrack_lock);
1051                 DEBUGP("no such conntrack, create new\n");
1052                 err = -ENOENT;
1053                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1054                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1055                 return err;
1056         }
1057         /* implicit 'else' */
1058
1059         /* we only allow nat config for new conntracks */
1060         if (cda[CTA_NAT-1]) {
1061                 err = -EINVAL;
1062                 goto out_unlock;
1063         }
1064
1065         /* We manipulate the conntrack inside the global conntrack table lock,
1066          * so there's no need to increase the refcount */
1067         DEBUGP("conntrack found\n");
1068         err = -EEXIST;
1069         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1070                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1071
1072 out_unlock:
1073         write_unlock_bh(&ip_conntrack_lock);
1074         return err;
1075 }
1076
1077 /***********************************************************************
1078  * EXPECT
1079  ***********************************************************************/
1080
1081 static inline int
1082 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1083                          const struct ip_conntrack_tuple *tuple,
1084                          enum ctattr_expect type)
1085 {
1086         struct nfattr *nest_parms = NFA_NEST(skb, type);
1087
1088         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1089                 goto nfattr_failure;
1090
1091         NFA_NEST_END(skb, nest_parms);
1092
1093         return 0;
1094
1095 nfattr_failure:
1096         return -1;
1097 }
1098
1099 static inline int
1100 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1101                           const struct ip_conntrack_expect *exp)
1102 {
1103         struct ip_conntrack *master = exp->master;
1104         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1105         u_int32_t id = htonl(exp->id);
1106
1107         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1108                 goto nfattr_failure;
1109         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1110                 goto nfattr_failure;
1111         if (ctnetlink_exp_dump_tuple(skb,
1112                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1113                                  CTA_EXPECT_MASTER) < 0)
1114                 goto nfattr_failure;
1115
1116         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1117         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1118
1119         return 0;
1120
1121 nfattr_failure:
1122         return -1;
1123 }
1124
1125 static int
1126 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1127                     int event,
1128                     int nowait,
1129                     const struct ip_conntrack_expect *exp)
1130 {
1131         struct nlmsghdr *nlh;
1132         struct nfgenmsg *nfmsg;
1133         unsigned char *b;
1134
1135         b = skb->tail;
1136
1137         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1138         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1139         nfmsg  = NLMSG_DATA(nlh);
1140
1141         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1142         nfmsg->nfgen_family = AF_INET;
1143         nfmsg->version      = NFNETLINK_V0;
1144         nfmsg->res_id       = 0;
1145
1146         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1147                 goto nfattr_failure;
1148
1149         nlh->nlmsg_len = skb->tail - b;
1150         return skb->len;
1151
1152 nlmsg_failure:
1153 nfattr_failure:
1154         skb_trim(skb, b - skb->data);
1155         return -1;
1156 }
1157
1158 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1159 static int ctnetlink_expect_event(struct notifier_block *this,
1160                                   unsigned long events, void *ptr)
1161 {
1162         struct nlmsghdr *nlh;
1163         struct nfgenmsg *nfmsg;
1164         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1165         struct sk_buff *skb;
1166         unsigned int type;
1167         unsigned char *b;
1168         int flags = 0;
1169         u16 proto;
1170
1171         if (events & IPEXP_NEW) {
1172                 type = IPCTNL_MSG_EXP_NEW;
1173                 flags = NLM_F_CREATE|NLM_F_EXCL;
1174         } else
1175                 return NOTIFY_DONE;
1176
1177         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1178         if (!skb)
1179                 return NOTIFY_DONE;
1180
1181         b = skb->tail;
1182
1183         type |= NFNL_SUBSYS_CTNETLINK << 8;
1184         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1185         nfmsg = NLMSG_DATA(nlh);
1186
1187         nlh->nlmsg_flags    = flags;
1188         nfmsg->nfgen_family = AF_INET;
1189         nfmsg->version      = NFNETLINK_V0;
1190         nfmsg->res_id       = 0;
1191
1192         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1193                 goto nfattr_failure;
1194
1195         nlh->nlmsg_len = skb->tail - b;
1196         proto = exp->tuple.dst.protonum;
1197         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1198         return NOTIFY_DONE;
1199
1200 nlmsg_failure:
1201 nfattr_failure:
1202         kfree_skb(skb);
1203         return NOTIFY_DONE;
1204 }
1205 #endif
1206
1207 static int
1208 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1209 {
1210         struct ip_conntrack_expect *exp = NULL;
1211         struct list_head *i;
1212         u_int32_t *id = (u_int32_t *) &cb->args[0];
1213
1214         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1215
1216         read_lock_bh(&ip_conntrack_lock);
1217         list_for_each_prev(i, &ip_conntrack_expect_list) {
1218                 exp = (struct ip_conntrack_expect *) i;
1219                 if (exp->id <= *id)
1220                         continue;
1221                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1222                                             cb->nlh->nlmsg_seq,
1223                                             IPCTNL_MSG_EXP_NEW,
1224                                             1, exp) < 0)
1225                         goto out;
1226                 *id = exp->id;
1227         }
1228 out:
1229         read_unlock_bh(&ip_conntrack_lock);
1230
1231         DEBUGP("leaving, last id=%llu\n", *id);
1232
1233         return skb->len;
1234 }
1235
1236 static int
1237 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1238                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1239 {
1240         struct ip_conntrack_tuple tuple;
1241         struct ip_conntrack_expect *exp;
1242         struct sk_buff *skb2;
1243         int err = 0;
1244
1245         DEBUGP("entered %s\n", __FUNCTION__);
1246
1247         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1248                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1249                 u32 rlen;
1250
1251                 if (msg->nfgen_family != AF_INET)
1252                         return -EAFNOSUPPORT;
1253
1254                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1255                                                 ctnetlink_exp_dump_table,
1256                                                 ctnetlink_done)) != 0)
1257                         return -EINVAL;
1258                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1259                 if (rlen > skb->len)
1260                         rlen = skb->len;
1261                 skb_pull(skb, rlen);
1262                 return 0;
1263         }
1264
1265         if (cda[CTA_EXPECT_MASTER-1])
1266                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1267         else
1268                 return -EINVAL;
1269
1270         if (err < 0)
1271                 return err;
1272
1273         exp = ip_conntrack_expect_find_get(&tuple);
1274         if (!exp)
1275                 return -ENOENT;
1276
1277         err = -ENOMEM;
1278         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1279         if (!skb2)
1280                 goto out;
1281         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1282
1283         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1284                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1285                                       1, exp);
1286         if (err <= 0)
1287                 goto out;
1288
1289         ip_conntrack_expect_put(exp);
1290
1291         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1292         if (err < 0)
1293                 goto free;
1294
1295         return err;
1296
1297 out:
1298         ip_conntrack_expect_put(exp);
1299 free:
1300         if (skb2)
1301                 kfree_skb(skb2);
1302         return err;
1303 }
1304
1305 static int
1306 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1307                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1308 {
1309         struct ip_conntrack_expect *exp, *tmp;
1310         struct ip_conntrack_tuple tuple;
1311         struct ip_conntrack_helper *h;
1312         int err;
1313
1314         if (cda[CTA_EXPECT_TUPLE-1]) {
1315                 /* delete a single expect by tuple */
1316                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1317                 if (err < 0)
1318                         return err;
1319
1320                 /* bump usage count to 2 */
1321                 exp = ip_conntrack_expect_find_get(&tuple);
1322                 if (!exp)
1323                         return -ENOENT;
1324
1325                 if (cda[CTA_EXPECT_ID-1]) {
1326                         u_int32_t id =
1327                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1328                         if (exp->id != ntohl(id)) {
1329                                 ip_conntrack_expect_put(exp);
1330                                 return -ENOENT;
1331                         }
1332                 }
1333
1334                 /* after list removal, usage count == 1 */
1335                 ip_conntrack_unexpect_related(exp);
1336                 /* have to put what we 'get' above.
1337                  * after this line usage count == 0 */
1338                 ip_conntrack_expect_put(exp);
1339         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1340                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1341
1342                 /* delete all expectations for this helper */
1343                 write_lock_bh(&ip_conntrack_lock);
1344                 h = __ip_conntrack_helper_find_byname(name);
1345                 if (!h) {
1346                         write_unlock_bh(&ip_conntrack_lock);
1347                         return -EINVAL;
1348                 }
1349                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1350                                          list) {
1351                         if (exp->master->helper == h
1352                             && del_timer(&exp->timeout))
1353                                 __ip_ct_expect_unlink_destroy(exp);
1354                 }
1355                 write_unlock(&ip_conntrack_lock);
1356         } else {
1357                 /* This basically means we have to flush everything*/
1358                 write_lock_bh(&ip_conntrack_lock);
1359                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1360                                          list) {
1361                         if (del_timer(&exp->timeout))
1362                                 __ip_ct_expect_unlink_destroy(exp);
1363                 }
1364                 write_unlock_bh(&ip_conntrack_lock);
1365         }
1366
1367         return 0;
1368 }
1369 static int
1370 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1371 {
1372         return -EOPNOTSUPP;
1373 }
1374
1375 static int
1376 ctnetlink_create_expect(struct nfattr *cda[])
1377 {
1378         struct ip_conntrack_tuple tuple, mask, master_tuple;
1379         struct ip_conntrack_tuple_hash *h = NULL;
1380         struct ip_conntrack_expect *exp;
1381         struct ip_conntrack *ct;
1382         int err = 0;
1383
1384         DEBUGP("entered %s\n", __FUNCTION__);
1385
1386         /* caller guarantees that those three CTA_EXPECT_* exist */
1387         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1388         if (err < 0)
1389                 return err;
1390         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1391         if (err < 0)
1392                 return err;
1393         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1394         if (err < 0)
1395                 return err;
1396
1397         /* Look for master conntrack of this expectation */
1398         h = ip_conntrack_find_get(&master_tuple, NULL);
1399         if (!h)
1400                 return -ENOENT;
1401         ct = tuplehash_to_ctrack(h);
1402
1403         if (!ct->helper) {
1404                 /* such conntrack hasn't got any helper, abort */
1405                 err = -EINVAL;
1406                 goto out;
1407         }
1408
1409         exp = ip_conntrack_expect_alloc(ct);
1410         if (!exp) {
1411                 err = -ENOMEM;
1412                 goto out;
1413         }
1414
1415         exp->expectfn = NULL;
1416         exp->master = ct;
1417         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1418         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1419
1420         err = ip_conntrack_expect_related(exp);
1421         ip_conntrack_expect_put(exp);
1422
1423 out:
1424         ip_conntrack_put(tuplehash_to_ctrack(h));
1425         return err;
1426 }
1427
1428 static int
1429 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1430                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1431 {
1432         struct ip_conntrack_tuple tuple;
1433         struct ip_conntrack_expect *exp;
1434         int err = 0;
1435
1436         DEBUGP("entered %s\n", __FUNCTION__);
1437
1438         if (!cda[CTA_EXPECT_TUPLE-1]
1439             || !cda[CTA_EXPECT_MASK-1]
1440             || !cda[CTA_EXPECT_MASTER-1])
1441                 return -EINVAL;
1442
1443         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1444         if (err < 0)
1445                 return err;
1446
1447         write_lock_bh(&ip_conntrack_lock);
1448         exp = __ip_conntrack_expect_find(&tuple);
1449
1450         if (!exp) {
1451                 write_unlock_bh(&ip_conntrack_lock);
1452                 err = -ENOENT;
1453                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1454                         err = ctnetlink_create_expect(cda);
1455                 return err;
1456         }
1457
1458         err = -EEXIST;
1459         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1460                 err = ctnetlink_change_expect(exp, cda);
1461         write_unlock_bh(&ip_conntrack_lock);
1462
1463         DEBUGP("leaving\n");
1464
1465         return err;
1466 }
1467
1468 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1469 static struct notifier_block ctnl_notifier = {
1470         .notifier_call  = ctnetlink_conntrack_event,
1471 };
1472
1473 static struct notifier_block ctnl_notifier_exp = {
1474         .notifier_call  = ctnetlink_expect_event,
1475 };
1476 #endif
1477
1478 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1479         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1480                                             .attr_count = CTA_MAX,
1481                                             .cap_required = CAP_NET_ADMIN },
1482         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1483                                             .attr_count = CTA_MAX,
1484                                             .cap_required = CAP_NET_ADMIN },
1485         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1486                                             .attr_count = CTA_MAX,
1487                                             .cap_required = CAP_NET_ADMIN },
1488         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1489                                             .attr_count = CTA_MAX,
1490                                             .cap_required = CAP_NET_ADMIN },
1491 };
1492
1493 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1494         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1495                                             .attr_count = CTA_EXPECT_MAX,
1496                                             .cap_required = CAP_NET_ADMIN },
1497         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1498                                             .attr_count = CTA_EXPECT_MAX,
1499                                             .cap_required = CAP_NET_ADMIN },
1500         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1501                                             .attr_count = CTA_EXPECT_MAX,
1502                                             .cap_required = CAP_NET_ADMIN },
1503 };
1504
1505 static struct nfnetlink_subsystem ctnl_subsys = {
1506         .name                           = "conntrack",
1507         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1508         .cb_count                       = IPCTNL_MSG_MAX,
1509         .cb                             = ctnl_cb,
1510 };
1511
1512 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1513         .name                           = "conntrack_expect",
1514         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1515         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1516         .cb                             = ctnl_exp_cb,
1517 };
1518
1519 static int __init ctnetlink_init(void)
1520 {
1521         int ret;
1522
1523         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1524         ret = nfnetlink_subsys_register(&ctnl_subsys);
1525         if (ret < 0) {
1526                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1527                 goto err_out;
1528         }
1529
1530         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1531         if (ret < 0) {
1532                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1533                 goto err_unreg_subsys;
1534         }
1535
1536 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1537         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1538         if (ret < 0) {
1539                 printk("ctnetlink_init: cannot register notifier.\n");
1540                 goto err_unreg_exp_subsys;
1541         }
1542
1543         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1544         if (ret < 0) {
1545                 printk("ctnetlink_init: cannot expect register notifier.\n");
1546                 goto err_unreg_notifier;
1547         }
1548 #endif
1549
1550         return 0;
1551
1552 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1553 err_unreg_notifier:
1554         ip_conntrack_unregister_notifier(&ctnl_notifier);
1555 err_unreg_exp_subsys:
1556         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1557 #endif
1558 err_unreg_subsys:
1559         nfnetlink_subsys_unregister(&ctnl_subsys);
1560 err_out:
1561         return ret;
1562 }
1563
1564 static void __exit ctnetlink_exit(void)
1565 {
1566         printk("ctnetlink: unregistering from nfnetlink.\n");
1567
1568 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1569         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1570         ip_conntrack_unregister_notifier(&ctnl_notifier);
1571 #endif
1572
1573         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1574         nfnetlink_subsys_unregister(&ctnl_subsys);
1575         return;
1576 }
1577
1578 module_init(ctnetlink_init);
1579 module_exit(ctnetlink_exit);