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
  32 #include <linux/netfilter.h>
  33 #include <linux/netfilter_ipv4/ip_conntrack.h>
  34 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  37 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  38
  39 #include <linux/netfilter/nfnetlink.h>
  40 #include <linux/netfilter/nfnetlink_conntrack.h>
  41
  42 MODULE_LICENSE("GPL");
  43
  44 static char __initdata version[] = "0.90";
  45
  46 #if 0
  47 #define DEBUGP printk
  48 #else
  49 #define DEBUGP(format, args...)
  50 #endif
  51
  52
  53 static inline int
  54 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  55                             const struct ip_conntrack_tuple *tuple)
  56 {
  57         struct ip_conntrack_protocol *proto;
  58         int ret = 0;
  59
  60         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  61
  62         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  63         if (likely(proto && proto->tuple_to_nfattr)) {
  64                 ret = proto->tuple_to_nfattr(skb, tuple);
  65                 ip_conntrack_proto_put(proto);
  66         }
  67
  68         return ret;
  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_int32_t tmp;
 179
 180         tmp = htonl(ct->counters[dir].packets);
 181         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 182
 183         tmp = htonl(ct->counters[dir].bytes);
 184         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_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 size_t 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         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 483
 484         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 485                 return -EINVAL;
 486
 487         if (!tb[CTA_IP_V4_SRC-1])
 488                 return -EINVAL;
 489         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 490
 491         if (!tb[CTA_IP_V4_DST-1])
 492                 return -EINVAL;
 493         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 494
 495         DEBUGP("leaving\n");
 496
 497         return 0;
 498 }
 499
 500 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 501         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 502         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 503         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 504         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 505         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 506         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 507 };
 508
 509 static inline int
 510 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 511                             struct ip_conntrack_tuple *tuple)
 512 {
 513         struct nfattr *tb[CTA_PROTO_MAX];
 514         struct ip_conntrack_protocol *proto;
 515         int ret = 0;
 516
 517         DEBUGP("entered %s\n", __FUNCTION__);
 518
 519         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 520
 521         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 522                 return -EINVAL;
 523
 524         if (!tb[CTA_PROTO_NUM-1])
 525                 return -EINVAL;
 526         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 527
 528         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 529
 530         if (likely(proto && proto->nfattr_to_tuple)) {
 531                 ret = proto->nfattr_to_tuple(tb, tuple);
 532                 ip_conntrack_proto_put(proto);
 533         }
 534
 535         return ret;
 536 }
 537
 538 static inline int
 539 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 540                       enum ctattr_tuple type)
 541 {
 542         struct nfattr *tb[CTA_TUPLE_MAX];
 543         int err;
 544
 545         DEBUGP("entered %s\n", __FUNCTION__);
 546
 547         memset(tuple, 0, sizeof(*tuple));
 548
 549         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 550
 551         if (!tb[CTA_TUPLE_IP-1])
 552                 return -EINVAL;
 553
 554         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 555         if (err < 0)
 556                 return err;
 557
 558         if (!tb[CTA_TUPLE_PROTO-1])
 559                 return -EINVAL;
 560
 561         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 562         if (err < 0)
 563                 return err;
 564
 565         /* orig and expect tuples get DIR_ORIGINAL */
 566         if (type == CTA_TUPLE_REPLY)
 567                 tuple->dst.dir = IP_CT_DIR_REPLY;
 568         else
 569                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 570
 571         DUMP_TUPLE(tuple);
 572
 573         DEBUGP("leaving\n");
 574
 575         return 0;
 576 }
 577
 578 #ifdef CONFIG_IP_NF_NAT_NEEDED
 579 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 580         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 581         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 582 };
 583
 584 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 585                                      const struct ip_conntrack *ct,
 586                                      struct ip_nat_range *range)
 587 {
 588         struct nfattr *tb[CTA_PROTONAT_MAX];
 589         struct ip_nat_protocol *npt;
 590
 591         DEBUGP("entered %s\n", __FUNCTION__);
 592
 593         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 594
 595         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 596                 return -EINVAL;
 597
 598         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 599         if (!npt)
 600                 return 0;
 601
 602         if (!npt->nfattr_to_range) {
 603                 ip_nat_proto_put(npt);
 604                 return 0;
 605         }
 606
 607         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 608         if (npt->nfattr_to_range(tb, range) > 0)
 609                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 610
 611         ip_nat_proto_put(npt);
 612
 613         DEBUGP("leaving\n");
 614         return 0;
 615 }
 616
 617 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 618         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 619         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 620 };
 621
 622 static inline int
 623 ctnetlink_parse_nat(struct nfattr *cda[],
 624                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 625 {
 626         struct nfattr *tb[CTA_NAT_MAX];
 627         int err;
 628
 629         DEBUGP("entered %s\n", __FUNCTION__);
 630
 631         memset(range, 0, sizeof(*range));
 632
 633         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 634
 635         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 636                 return -EINVAL;
 637
 638         if (tb[CTA_NAT_MINIP-1])
 639                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 640
 641         if (!tb[CTA_NAT_MAXIP-1])
 642                 range->max_ip = range->min_ip;
 643         else
 644                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 645
 646         if (range->min_ip)
 647                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 648
 649         if (!tb[CTA_NAT_PROTO-1])
 650                 return 0;
 651
 652         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 653         if (err < 0)
 654                 return err;
 655
 656         DEBUGP("leaving\n");
 657         return 0;
 658 }
 659 #endif
 660
 661 static inline int
 662 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 663 {
 664         struct nfattr *tb[CTA_HELP_MAX];
 665
 666         DEBUGP("entered %s\n", __FUNCTION__);
 667
 668         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 669
 670         if (!tb[CTA_HELP_NAME-1])
 671                 return -EINVAL;
 672
 673         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 674
 675         return 0;
 676 }
 677
 678 static const size_t cta_min[CTA_MAX] = {
 679         [CTA_STATUS-1]          = sizeof(u_int32_t),
 680         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 681         [CTA_MARK-1]            = sizeof(u_int32_t),
 682         [CTA_USE-1]             = sizeof(u_int32_t),
 683         [CTA_ID-1]              = sizeof(u_int32_t)
 684 };
 685
 686 static int
 687 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 688                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 689 {
 690         struct ip_conntrack_tuple_hash *h;
 691         struct ip_conntrack_tuple tuple;
 692         struct ip_conntrack *ct;
 693         int err = 0;
 694
 695         DEBUGP("entered %s\n", __FUNCTION__);
 696
 697         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 698                 return -EINVAL;
 699
 700         if (cda[CTA_TUPLE_ORIG-1])
 701                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 702         else if (cda[CTA_TUPLE_REPLY-1])
 703                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 704         else {
 705                 /* Flush the whole table */
 706                 ip_conntrack_flush();
 707                 return 0;
 708         }
 709
 710         if (err < 0)
 711                 return err;
 712
 713         h = ip_conntrack_find_get(&tuple, NULL);
 714         if (!h) {
 715                 DEBUGP("tuple not found in conntrack hash\n");
 716                 return -ENOENT;
 717         }
 718
 719         ct = tuplehash_to_ctrack(h);
 720
 721         if (cda[CTA_ID-1]) {
 722                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 723                 if (ct->id != id) {
 724                         ip_conntrack_put(ct);
 725                         return -ENOENT;
 726                 }
 727         }
 728         if (del_timer(&ct->timeout)) {
 729                 ip_conntrack_put(ct);
 730                 ct->timeout.function((unsigned long)ct);
 731                 return 0;
 732         }
 733         ip_conntrack_put(ct);
 734         DEBUGP("leaving\n");
 735
 736         return 0;
 737 }
 738
 739 static int
 740 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 741                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 742 {
 743         struct ip_conntrack_tuple_hash *h;
 744         struct ip_conntrack_tuple tuple;
 745         struct ip_conntrack *ct;
 746         struct sk_buff *skb2 = NULL;
 747         int err = 0;
 748
 749         DEBUGP("entered %s\n", __FUNCTION__);
 750
 751         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 752                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 753                 u32 rlen;
 754
 755                 if (msg->nfgen_family != AF_INET)
 756                         return -EAFNOSUPPORT;
 757
 758                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 759                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 760 #ifdef CONFIG_IP_NF_CT_ACCT
 761                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 762                                                 ctnetlink_dump_table_w,
 763                                                 ctnetlink_done)) != 0)
 764                                 return -EINVAL;
 765 #else
 766                         return -ENOTSUPP;
 767 #endif
 768                 } else {
 769                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 770                                                         ctnetlink_dump_table,
 771                                                         ctnetlink_done)) != 0)
 772                         return -EINVAL;
 773                 }
 774
 775                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 776                 if (rlen > skb->len)
 777                         rlen = skb->len;
 778                 skb_pull(skb, rlen);
 779                 return 0;
 780         }
 781
 782         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 783                 return -EINVAL;
 784
 785         if (cda[CTA_TUPLE_ORIG-1])
 786                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 787         else if (cda[CTA_TUPLE_REPLY-1])
 788                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 789         else
 790                 return -EINVAL;
 791
 792         if (err < 0)
 793                 return err;
 794
 795         h = ip_conntrack_find_get(&tuple, NULL);
 796         if (!h) {
 797                 DEBUGP("tuple not found in conntrack hash");
 798                 return -ENOENT;
 799         }
 800         DEBUGP("tuple found\n");
 801         ct = tuplehash_to_ctrack(h);
 802
 803         err = -ENOMEM;
 804         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 805         if (!skb2) {
 806                 ip_conntrack_put(ct);
 807                 return -ENOMEM;
 808         }
 809         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 810
 811         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 812                                   IPCTNL_MSG_CT_NEW, 1, ct);
 813         ip_conntrack_put(ct);
 814         if (err <= 0)
 815                 goto free;
 816
 817         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 818         if (err < 0)
 819                 goto out;
 820
 821         DEBUGP("leaving\n");
 822         return 0;
 823
 824 free:
 825         kfree_skb(skb2);
 826 out:
 827         return err;
 828 }
 829
 830 static inline int
 831 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 832 {
 833         unsigned long d;
 834         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 835         d = ct->status ^ status;
 836
 837         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 838                 /* unchangeable */
 839                 return -EINVAL;
 840
 841         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 842                 /* SEEN_REPLY bit can only be set */
 843                 return -EINVAL;
 844
 845
 846         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 847                 /* ASSURED bit can only be set */
 848                 return -EINVAL;
 849
 850         if (cda[CTA_NAT-1]) {
 851 #ifndef CONFIG_IP_NF_NAT_NEEDED
 852                 return -EINVAL;
 853 #else
 854                 unsigned int hooknum;
 855                 struct ip_nat_range range;
 856
 857                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 858                         return -EINVAL;
 859
 860                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 861                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 862                        htons(range.min.all), htons(range.max.all));
 863
 864                 /* This is tricky but it works. ip_nat_setup_info needs the
 865                  * hook number as parameter, so let's do the correct
 866                  * conversion and run away */
 867                 if (status & IPS_SRC_NAT_DONE)
 868                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 869                 else if (status & IPS_DST_NAT_DONE)
 870                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 871                 else
 872                         return -EINVAL; /* Missing NAT flags */
 873
 874                 DEBUGP("NAT status: %lu\n",
 875                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 876
 877                 if (ip_nat_initialized(ct, hooknum))
 878                         return -EEXIST;
 879                 ip_nat_setup_info(ct, &range, hooknum);
 880
 881                 DEBUGP("NAT status after setup_info: %lu\n",
 882                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 883 #endif
 884         }
 885
 886         /* Be careful here, modifying NAT bits can screw up things,
 887          * so don't let users modify them directly if they don't pass
 888          * ip_nat_range. */
 889         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 890         return 0;
 891 }
 892
 893
 894 static inline int
 895 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 896 {
 897         struct ip_conntrack_helper *helper;
 898         char *helpname;
 899         int err;
 900
 901         DEBUGP("entered %s\n", __FUNCTION__);
 902
 903         /* don't change helper of sibling connections */
 904         if (ct->master)
 905                 return -EINVAL;
 906
 907         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 908         if (err < 0)
 909                 return err;
 910
 911         helper = __ip_conntrack_helper_find_byname(helpname);
 912         if (!helper) {
 913                 if (!strcmp(helpname, ""))
 914                         helper = NULL;
 915                 else
 916                         return -EINVAL;
 917         }
 918
 919         if (ct->helper) {
 920                 if (!helper) {
 921                         /* we had a helper before ... */
 922                         ip_ct_remove_expectations(ct);
 923                         ct->helper = NULL;
 924                 } else {
 925                         /* need to zero data of old helper */
 926                         memset(&ct->help, 0, sizeof(ct->help));
 927                 }
 928         }
 929
 930         ct->helper = helper;
 931
 932         return 0;
 933 }
 934
 935 static inline int
 936 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 937 {
 938         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 939
 940         if (!del_timer(&ct->timeout))
 941                 return -ETIME;
 942
 943         ct->timeout.expires = jiffies + timeout * HZ;
 944         add_timer(&ct->timeout);
 945
 946         return 0;
 947 }
 948
 949 static inline int
 950 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 951 {
 952         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 953         struct ip_conntrack_protocol *proto;
 954         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 955         int err = 0;
 956
 957         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 958
 959         proto = ip_conntrack_proto_find_get(npt);
 960         if (!proto)
 961                 return -EINVAL;
 962
 963         if (proto->from_nfattr)
 964                 err = proto->from_nfattr(tb, ct);
 965         ip_conntrack_proto_put(proto);
 966
 967         return err;
 968 }
 969
 970 static int
 971 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 972 {
 973         int err;
 974
 975         DEBUGP("entered %s\n", __FUNCTION__);
 976
 977         if (cda[CTA_HELP-1]) {
 978                 err = ctnetlink_change_helper(ct, cda);
 979                 if (err < 0)
 980                         return err;
 981         }
 982
 983         if (cda[CTA_TIMEOUT-1]) {
 984                 err = ctnetlink_change_timeout(ct, cda);
 985                 if (err < 0)
 986                         return err;
 987         }
 988
 989         if (cda[CTA_STATUS-1]) {
 990                 err = ctnetlink_change_status(ct, cda);
 991                 if (err < 0)
 992                         return err;
 993         }
 994
 995         if (cda[CTA_PROTOINFO-1]) {
 996                 err = ctnetlink_change_protoinfo(ct, cda);
 997                 if (err < 0)
 998                         return err;
 999         }
1000
1001 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1002         if (cda[CTA_MARK-1])
1003                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1004 #endif
1005
1006         DEBUGP("all done\n");
1007         return 0;
1008 }
1009
1010 static int
1011 ctnetlink_create_conntrack(struct nfattr *cda[],
1012                            struct ip_conntrack_tuple *otuple,
1013                            struct ip_conntrack_tuple *rtuple)
1014 {
1015         struct ip_conntrack *ct;
1016         int err = -EINVAL;
1017
1018         DEBUGP("entered %s\n", __FUNCTION__);
1019
1020         ct = ip_conntrack_alloc(otuple, rtuple);
1021         if (ct == NULL || IS_ERR(ct))
1022                 return -ENOMEM;
1023
1024         if (!cda[CTA_TIMEOUT-1])
1025                 goto err;
1026         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1027
1028         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1029         ct->status |= IPS_CONFIRMED;
1030
1031         err = ctnetlink_change_status(ct, cda);
1032         if (err < 0)
1033                 goto err;
1034
1035         if (cda[CTA_PROTOINFO-1]) {
1036                 err = ctnetlink_change_protoinfo(ct, cda);
1037                 if (err < 0)
1038                         return err;
1039         }
1040
1041         ct->helper = ip_conntrack_helper_find_get(rtuple);
1042
1043         add_timer(&ct->timeout);
1044         ip_conntrack_hash_insert(ct);
1045
1046         if (ct->helper)
1047                 ip_conntrack_helper_put(ct->helper);
1048
1049 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1050         if (cda[CTA_MARK-1])
1051                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1052 #endif
1053
1054         DEBUGP("conntrack with id %u inserted\n", ct->id);
1055         return 0;
1056
1057 err:
1058         ip_conntrack_free(ct);
1059         return err;
1060 }
1061
1062 static int
1063 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1064                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1065 {
1066         struct ip_conntrack_tuple otuple, rtuple;
1067         struct ip_conntrack_tuple_hash *h = NULL;
1068         int err = 0;
1069
1070         DEBUGP("entered %s\n", __FUNCTION__);
1071
1072         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1073                 return -EINVAL;
1074
1075         if (cda[CTA_TUPLE_ORIG-1]) {
1076                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1077                 if (err < 0)
1078                         return err;
1079         }
1080
1081         if (cda[CTA_TUPLE_REPLY-1]) {
1082                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1083                 if (err < 0)
1084                         return err;
1085         }
1086
1087         write_lock_bh(&ip_conntrack_lock);
1088         if (cda[CTA_TUPLE_ORIG-1])
1089                 h = __ip_conntrack_find(&otuple, NULL);
1090         else if (cda[CTA_TUPLE_REPLY-1])
1091                 h = __ip_conntrack_find(&rtuple, NULL);
1092
1093         if (h == NULL) {
1094                 write_unlock_bh(&ip_conntrack_lock);
1095                 DEBUGP("no such conntrack, create new\n");
1096                 err = -ENOENT;
1097                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1098                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1099                 return err;
1100         }
1101         /* implicit 'else' */
1102
1103         /* we only allow nat config for new conntracks */
1104         if (cda[CTA_NAT-1]) {
1105                 err = -EINVAL;
1106                 goto out_unlock;
1107         }
1108
1109         /* We manipulate the conntrack inside the global conntrack table lock,
1110          * so there's no need to increase the refcount */
1111         DEBUGP("conntrack found\n");
1112         err = -EEXIST;
1113         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1114                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1115
1116 out_unlock:
1117         write_unlock_bh(&ip_conntrack_lock);
1118         return err;
1119 }
1120
1121 /***********************************************************************
1122  * EXPECT
1123  ***********************************************************************/
1124
1125 static inline int
1126 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1127                          const struct ip_conntrack_tuple *tuple,
1128                          enum ctattr_expect type)
1129 {
1130         struct nfattr *nest_parms = NFA_NEST(skb, type);
1131
1132         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1133                 goto nfattr_failure;
1134
1135         NFA_NEST_END(skb, nest_parms);
1136
1137         return 0;
1138
1139 nfattr_failure:
1140         return -1;
1141 }
1142
1143 static inline int
1144 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1145                           const struct ip_conntrack_expect *exp)
1146 {
1147         struct ip_conntrack *master = exp->master;
1148         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1149         u_int32_t id = htonl(exp->id);
1150
1151         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1152                 goto nfattr_failure;
1153         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1154                 goto nfattr_failure;
1155         if (ctnetlink_exp_dump_tuple(skb,
1156                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1157                                  CTA_EXPECT_MASTER) < 0)
1158                 goto nfattr_failure;
1159
1160         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1161         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1162
1163         return 0;
1164
1165 nfattr_failure:
1166         return -1;
1167 }
1168
1169 static int
1170 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1171                     int event,
1172                     int nowait,
1173                     const struct ip_conntrack_expect *exp)
1174 {
1175         struct nlmsghdr *nlh;
1176         struct nfgenmsg *nfmsg;
1177         unsigned char *b;
1178
1179         b = skb->tail;
1180
1181         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1182         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1183         nfmsg  = NLMSG_DATA(nlh);
1184
1185         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1186         nfmsg->nfgen_family = AF_INET;
1187         nfmsg->version      = NFNETLINK_V0;
1188         nfmsg->res_id       = 0;
1189
1190         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1191                 goto nfattr_failure;
1192
1193         nlh->nlmsg_len = skb->tail - b;
1194         return skb->len;
1195
1196 nlmsg_failure:
1197 nfattr_failure:
1198         skb_trim(skb, b - skb->data);
1199         return -1;
1200 }
1201
1202 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1203 static int ctnetlink_expect_event(struct notifier_block *this,
1204                                   unsigned long events, void *ptr)
1205 {
1206         struct nlmsghdr *nlh;
1207         struct nfgenmsg *nfmsg;
1208         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1209         struct sk_buff *skb;
1210         unsigned int type;
1211         unsigned char *b;
1212         int flags = 0;
1213         u16 proto;
1214
1215         if (events & IPEXP_NEW) {
1216                 type = IPCTNL_MSG_EXP_NEW;
1217                 flags = NLM_F_CREATE|NLM_F_EXCL;
1218         } else
1219                 return NOTIFY_DONE;
1220
1221         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1222         if (!skb)
1223                 return NOTIFY_DONE;
1224
1225         b = skb->tail;
1226
1227         type |= NFNL_SUBSYS_CTNETLINK << 8;
1228         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1229         nfmsg = NLMSG_DATA(nlh);
1230
1231         nlh->nlmsg_flags    = flags;
1232         nfmsg->nfgen_family = AF_INET;
1233         nfmsg->version      = NFNETLINK_V0;
1234         nfmsg->res_id       = 0;
1235
1236         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1237                 goto nfattr_failure;
1238
1239         nlh->nlmsg_len = skb->tail - b;
1240         proto = exp->tuple.dst.protonum;
1241         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1242         return NOTIFY_DONE;
1243
1244 nlmsg_failure:
1245 nfattr_failure:
1246         kfree_skb(skb);
1247         return NOTIFY_DONE;
1248 }
1249 #endif
1250
1251 static int
1252 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1253 {
1254         struct ip_conntrack_expect *exp = NULL;
1255         struct list_head *i;
1256         u_int32_t *id = (u_int32_t *) &cb->args[0];
1257
1258         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1259
1260         read_lock_bh(&ip_conntrack_lock);
1261         list_for_each_prev(i, &ip_conntrack_expect_list) {
1262                 exp = (struct ip_conntrack_expect *) i;
1263                 if (exp->id <= *id)
1264                         continue;
1265                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1266                                             cb->nlh->nlmsg_seq,
1267                                             IPCTNL_MSG_EXP_NEW,
1268                                             1, exp) < 0)
1269                         goto out;
1270                 *id = exp->id;
1271         }
1272 out:
1273         read_unlock_bh(&ip_conntrack_lock);
1274
1275         DEBUGP("leaving, last id=%llu\n", *id);
1276
1277         return skb->len;
1278 }
1279
1280 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1281         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1282         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1283 };
1284
1285 static int
1286 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1287                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1288 {
1289         struct ip_conntrack_tuple tuple;
1290         struct ip_conntrack_expect *exp;
1291         struct sk_buff *skb2;
1292         int err = 0;
1293
1294         DEBUGP("entered %s\n", __FUNCTION__);
1295
1296         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1297                 return -EINVAL;
1298
1299         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1300                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1301                 u32 rlen;
1302
1303                 if (msg->nfgen_family != AF_INET)
1304                         return -EAFNOSUPPORT;
1305
1306                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1307                                                 ctnetlink_exp_dump_table,
1308                                                 ctnetlink_done)) != 0)
1309                         return -EINVAL;
1310                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1311                 if (rlen > skb->len)
1312                         rlen = skb->len;
1313                 skb_pull(skb, rlen);
1314                 return 0;
1315         }
1316
1317         if (cda[CTA_EXPECT_MASTER-1])
1318                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1319         else
1320                 return -EINVAL;
1321
1322         if (err < 0)
1323                 return err;
1324
1325         exp = ip_conntrack_expect_find(&tuple);
1326         if (!exp)
1327                 return -ENOENT;
1328
1329         if (cda[CTA_EXPECT_ID-1]) {
1330                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1331                 if (exp->id != ntohl(id)) {
1332                         ip_conntrack_expect_put(exp);
1333                         return -ENOENT;
1334                 }
1335         }
1336
1337         err = -ENOMEM;
1338         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1339         if (!skb2)
1340                 goto out;
1341         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1342
1343         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1344                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1345                                       1, exp);
1346         if (err <= 0)
1347                 goto free;
1348
1349         ip_conntrack_expect_put(exp);
1350
1351         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1352
1353 free:
1354         kfree_skb(skb2);
1355 out:
1356         ip_conntrack_expect_put(exp);
1357         return err;
1358 }
1359
1360 static int
1361 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1362                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1363 {
1364         struct ip_conntrack_expect *exp, *tmp;
1365         struct ip_conntrack_tuple tuple;
1366         struct ip_conntrack_helper *h;
1367         int err;
1368
1369         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1370                 return -EINVAL;
1371
1372         if (cda[CTA_EXPECT_TUPLE-1]) {
1373                 /* delete a single expect by tuple */
1374                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1375                 if (err < 0)
1376                         return err;
1377
1378                 /* bump usage count to 2 */
1379                 exp = ip_conntrack_expect_find(&tuple);
1380                 if (!exp)
1381                         return -ENOENT;
1382
1383                 if (cda[CTA_EXPECT_ID-1]) {
1384                         u_int32_t id =
1385                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1386                         if (exp->id != ntohl(id)) {
1387                                 ip_conntrack_expect_put(exp);
1388                                 return -ENOENT;
1389                         }
1390                 }
1391
1392                 /* after list removal, usage count == 1 */
1393                 ip_conntrack_unexpect_related(exp);
1394                 /* have to put what we 'get' above.
1395                  * after this line usage count == 0 */
1396                 ip_conntrack_expect_put(exp);
1397         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1398                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1399
1400                 /* delete all expectations for this helper */
1401                 write_lock_bh(&ip_conntrack_lock);
1402                 h = __ip_conntrack_helper_find_byname(name);
1403                 if (!h) {
1404                         write_unlock_bh(&ip_conntrack_lock);
1405                         return -EINVAL;
1406                 }
1407                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1408                                          list) {
1409                         if (exp->master->helper == h
1410                             && del_timer(&exp->timeout)) {
1411                                 ip_ct_unlink_expect(exp);
1412                                 ip_conntrack_expect_put(exp);
1413                         }
1414                 }
1415                 write_unlock_bh(&ip_conntrack_lock);
1416         } else {
1417                 /* This basically means we have to flush everything*/
1418                 write_lock_bh(&ip_conntrack_lock);
1419                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1420                                          list) {
1421                         if (del_timer(&exp->timeout)) {
1422                                 ip_ct_unlink_expect(exp);
1423                                 ip_conntrack_expect_put(exp);
1424                         }
1425                 }
1426                 write_unlock_bh(&ip_conntrack_lock);
1427         }
1428
1429         return 0;
1430 }
1431 static int
1432 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1433 {
1434         return -EOPNOTSUPP;
1435 }
1436
1437 static int
1438 ctnetlink_create_expect(struct nfattr *cda[])
1439 {
1440         struct ip_conntrack_tuple tuple, mask, master_tuple;
1441         struct ip_conntrack_tuple_hash *h = NULL;
1442         struct ip_conntrack_expect *exp;
1443         struct ip_conntrack *ct;
1444         int err = 0;
1445
1446         DEBUGP("entered %s\n", __FUNCTION__);
1447
1448         /* caller guarantees that those three CTA_EXPECT_* exist */
1449         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1450         if (err < 0)
1451                 return err;
1452         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1453         if (err < 0)
1454                 return err;
1455         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1456         if (err < 0)
1457                 return err;
1458
1459         /* Look for master conntrack of this expectation */
1460         h = ip_conntrack_find_get(&master_tuple, NULL);
1461         if (!h)
1462                 return -ENOENT;
1463         ct = tuplehash_to_ctrack(h);
1464
1465         if (!ct->helper) {
1466                 /* such conntrack hasn't got any helper, abort */
1467                 err = -EINVAL;
1468                 goto out;
1469         }
1470
1471         exp = ip_conntrack_expect_alloc(ct);
1472         if (!exp) {
1473                 err = -ENOMEM;
1474                 goto out;
1475         }
1476
1477         exp->expectfn = NULL;
1478         exp->flags = 0;
1479         exp->master = ct;
1480         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1481         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1482
1483         err = ip_conntrack_expect_related(exp);
1484         ip_conntrack_expect_put(exp);
1485
1486 out:
1487         ip_conntrack_put(tuplehash_to_ctrack(h));
1488         return err;
1489 }
1490
1491 static int
1492 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1493                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1494 {
1495         struct ip_conntrack_tuple tuple;
1496         struct ip_conntrack_expect *exp;
1497         int err = 0;
1498
1499         DEBUGP("entered %s\n", __FUNCTION__);
1500
1501         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1502                 return -EINVAL;
1503
1504         if (!cda[CTA_EXPECT_TUPLE-1]
1505             || !cda[CTA_EXPECT_MASK-1]
1506             || !cda[CTA_EXPECT_MASTER-1])
1507                 return -EINVAL;
1508
1509         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1510         if (err < 0)
1511                 return err;
1512
1513         write_lock_bh(&ip_conntrack_lock);
1514         exp = __ip_conntrack_expect_find(&tuple);
1515
1516         if (!exp) {
1517                 write_unlock_bh(&ip_conntrack_lock);
1518                 err = -ENOENT;
1519                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1520                         err = ctnetlink_create_expect(cda);
1521                 return err;
1522         }
1523
1524         err = -EEXIST;
1525         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1526                 err = ctnetlink_change_expect(exp, cda);
1527         write_unlock_bh(&ip_conntrack_lock);
1528
1529         DEBUGP("leaving\n");
1530
1531         return err;
1532 }
1533
1534 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1535 static struct notifier_block ctnl_notifier = {
1536         .notifier_call  = ctnetlink_conntrack_event,
1537 };
1538
1539 static struct notifier_block ctnl_notifier_exp = {
1540         .notifier_call  = ctnetlink_expect_event,
1541 };
1542 #endif
1543
1544 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1545         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1546                                             .attr_count = CTA_MAX,
1547                                             .cap_required = CAP_NET_ADMIN },
1548         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1549                                             .attr_count = CTA_MAX,
1550                                             .cap_required = CAP_NET_ADMIN },
1551         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1552                                             .attr_count = CTA_MAX,
1553                                             .cap_required = CAP_NET_ADMIN },
1554         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1555                                             .attr_count = CTA_MAX,
1556                                             .cap_required = CAP_NET_ADMIN },
1557 };
1558
1559 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1560         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1561                                             .attr_count = CTA_EXPECT_MAX,
1562                                             .cap_required = CAP_NET_ADMIN },
1563         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1564                                             .attr_count = CTA_EXPECT_MAX,
1565                                             .cap_required = CAP_NET_ADMIN },
1566         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1567                                             .attr_count = CTA_EXPECT_MAX,
1568                                             .cap_required = CAP_NET_ADMIN },
1569 };
1570
1571 static struct nfnetlink_subsystem ctnl_subsys = {
1572         .name                           = "conntrack",
1573         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1574         .cb_count                       = IPCTNL_MSG_MAX,
1575         .cb                             = ctnl_cb,
1576 };
1577
1578 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1579         .name                           = "conntrack_expect",
1580         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1581         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1582         .cb                             = ctnl_exp_cb,
1583 };
1584
1585 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1586
1587 static int __init ctnetlink_init(void)
1588 {
1589         int ret;
1590
1591         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1592         ret = nfnetlink_subsys_register(&ctnl_subsys);
1593         if (ret < 0) {
1594                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1595                 goto err_out;
1596         }
1597
1598         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1599         if (ret < 0) {
1600                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1601                 goto err_unreg_subsys;
1602         }
1603
1604 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1605         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1606         if (ret < 0) {
1607                 printk("ctnetlink_init: cannot register notifier.\n");
1608                 goto err_unreg_exp_subsys;
1609         }
1610
1611         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1612         if (ret < 0) {
1613                 printk("ctnetlink_init: cannot expect register notifier.\n");
1614                 goto err_unreg_notifier;
1615         }
1616 #endif
1617
1618         return 0;
1619
1620 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1621 err_unreg_notifier:
1622         ip_conntrack_unregister_notifier(&ctnl_notifier);
1623 err_unreg_exp_subsys:
1624         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1625 #endif
1626 err_unreg_subsys:
1627         nfnetlink_subsys_unregister(&ctnl_subsys);
1628 err_out:
1629         return ret;
1630 }
1631
1632 static void __exit ctnetlink_exit(void)
1633 {
1634         printk("ctnetlink: unregistering from nfnetlink.\n");
1635
1636 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1637         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1638         ip_conntrack_unregister_notifier(&ctnl_notifier);
1639 #endif
1640
1641         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1642         nfnetlink_subsys_unregister(&ctnl_subsys);
1643         return;
1644 }
1645
1646 module_init(ctnetlink_init);
1647 module_exit(ctnetlink_exit);