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