SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv4/netfilter/ip_conntrack_netlink.c

   1 /* Connection tracking via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
   7  * (C) 2005-2006 by Pablo Neira Ayuso <pablo@eurodev.net>
   8  *
   9  * I've reworked this stuff to use attributes instead of conntrack
  10  * structures. 5.44 am. I need more tea. --pablo 05/07/11.
  11  *
  12  * Initial connection tracking via netlink development funded and
  13  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
  14  *
  15  * Further development of this code funded by Astaro AG (http://www.astaro.com)
  16  *
  17  * This software may be used and distributed according to the terms
  18  * of the GNU General Public License, incorporated herein by reference.
  19  */
  20
  21 #include <linux/init.h>
  22 #include <linux/module.h>
  23 #include <linux/kernel.h>
  24 #include <linux/types.h>
  25 #include <linux/timer.h>
  26 #include <linux/skbuff.h>
  27 #include <linux/errno.h>
  28 #include <linux/netlink.h>
  29 #include <linux/spinlock.h>
  30 #include <linux/interrupt.h>
  31 #include <linux/notifier.h>
  32
  33 #include <linux/netfilter.h>
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  38 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  39
  40 #include <linux/netfilter/nfnetlink.h>
  41 #include <linux/netfilter/nfnetlink_conntrack.h>
  42
  43 MODULE_LICENSE("GPL");
  44
  45 static char __initdata version[] = "0.90";
  46
  47 #if 0
  48 #define DEBUGP printk
  49 #else
  50 #define DEBUGP(format, args...)
  51 #endif
  52
  53
  54 static inline int
  55 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  56                             const struct ip_conntrack_tuple *tuple,
  57                             struct ip_conntrack_protocol *proto)
  58 {
  59         int ret = 0;
  60         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  61
  62         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  63
  64         if (likely(proto->tuple_to_nfattr))
  65                 ret = proto->tuple_to_nfattr(skb, tuple);
  66
  67         NFA_NEST_END(skb, nest_parms);
  68
  69         return ret;
  70
  71 nfattr_failure:
  72         return -1;
  73 }
  74
  75 static inline int
  76 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
  77                          const struct ip_conntrack_tuple *tuple)
  78 {
  79         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  80
  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
  84         NFA_NEST_END(skb, nest_parms);
  85
  86         return 0;
  87
  88 nfattr_failure:
  89         return -1;
  90 }
  91
  92 static inline int
  93 ctnetlink_dump_tuples(struct sk_buff *skb,
  94                       const struct ip_conntrack_tuple *tuple)
  95 {
  96         int ret;
  97         struct ip_conntrack_protocol *proto;
  98
  99         ret = ctnetlink_dump_tuples_ip(skb, tuple);
 100         if (unlikely(ret < 0))
 101                 return ret;
 102
 103         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 104         ret = ctnetlink_dump_tuples_proto(skb, tuple, proto);
 105         ip_conntrack_proto_put(proto);
 106
 107         return ret;
 108 }
 109
 110 static inline int
 111 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 112 {
 113         u_int32_t status = htonl((u_int32_t) ct->status);
 114         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 115         return 0;
 116
 117 nfattr_failure:
 118         return -1;
 119 }
 120
 121 static inline int
 122 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 123 {
 124         long timeout_l = ct->timeout.expires - jiffies;
 125         u_int32_t timeout;
 126
 127         if (timeout_l < 0)
 128                 timeout = 0;
 129         else
 130                 timeout = htonl(timeout_l / HZ);
 131
 132         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 133         return 0;
 134
 135 nfattr_failure:
 136         return -1;
 137 }
 138
 139 static inline int
 140 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 141 {
 142         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 143
 144         struct nfattr *nest_proto;
 145         int ret;
 146
 147         if (!proto->to_nfattr) {
 148                 ip_conntrack_proto_put(proto);
 149                 return 0;
 150         }
 151
 152         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 153
 154         ret = proto->to_nfattr(skb, nest_proto, ct);
 155
 156         ip_conntrack_proto_put(proto);
 157
 158         NFA_NEST_END(skb, nest_proto);
 159
 160         return ret;
 161
 162 nfattr_failure:
 163         return -1;
 164 }
 165
 166 static inline int
 167 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 168 {
 169         struct nfattr *nest_helper;
 170
 171         if (!ct->helper)
 172                 return 0;
 173
 174         nest_helper = NFA_NEST(skb, CTA_HELP);
 175         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 176
 177         if (ct->helper->to_nfattr)
 178                 ct->helper->to_nfattr(skb, ct);
 179
 180         NFA_NEST_END(skb, nest_helper);
 181
 182         return 0;
 183
 184 nfattr_failure:
 185         return -1;
 186 }
 187
 188 #ifdef CONFIG_IP_NF_CT_ACCT
 189 static inline int
 190 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 191                         enum ip_conntrack_dir dir)
 192 {
 193         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 194         struct nfattr *nest_count = NFA_NEST(skb, type);
 195         u_int32_t tmp;
 196
 197         tmp = htonl(ct->counters[dir].packets);
 198         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 199
 200         tmp = htonl(ct->counters[dir].bytes);
 201         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 202
 203         NFA_NEST_END(skb, nest_count);
 204
 205         return 0;
 206
 207 nfattr_failure:
 208         return -1;
 209 }
 210 #else
 211 #define ctnetlink_dump_counters(a, b, c) (0)
 212 #endif
 213
 214 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 215 static inline int
 216 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 217 {
 218         u_int32_t mark = htonl(ct->mark);
 219
 220         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 221         return 0;
 222
 223 nfattr_failure:
 224         return -1;
 225 }
 226 #else
 227 #define ctnetlink_dump_mark(a, b) (0)
 228 #endif
 229
 230 static inline int
 231 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 232 {
 233         u_int32_t id = htonl(ct->id);
 234         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 235         return 0;
 236
 237 nfattr_failure:
 238         return -1;
 239 }
 240
 241 static inline int
 242 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 243 {
 244         u_int32_t use = htonl(atomic_read(&ct->ct_general.use));
 245
 246         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 247         return 0;
 248
 249 nfattr_failure:
 250         return -1;
 251 }
 252
 253 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 254
 255 static int
 256 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 257                     int event, int nowait,
 258                     const struct ip_conntrack *ct)
 259 {
 260         struct nlmsghdr *nlh;
 261         struct nfgenmsg *nfmsg;
 262         struct nfattr *nest_parms;
 263         unsigned char *b;
 264
 265         b = skb->tail;
 266
 267         event |= NFNL_SUBSYS_CTNETLINK << 8;
 268         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 269         nfmsg  = NLMSG_DATA(nlh);
 270
 271         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 272         nfmsg->nfgen_family = AF_INET;
 273         nfmsg->version      = NFNETLINK_V0;
 274         nfmsg->res_id       = 0;
 275
 276         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 277         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 278                 goto nfattr_failure;
 279         NFA_NEST_END(skb, nest_parms);
 280
 281         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 282         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 283                 goto nfattr_failure;
 284         NFA_NEST_END(skb, nest_parms);
 285
 286         if (ctnetlink_dump_status(skb, ct) < 0 ||
 287             ctnetlink_dump_timeout(skb, ct) < 0 ||
 288             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 289             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 290             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 291             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 292             ctnetlink_dump_mark(skb, ct) < 0 ||
 293             ctnetlink_dump_id(skb, ct) < 0 ||
 294             ctnetlink_dump_use(skb, ct) < 0)
 295                 goto nfattr_failure;
 296
 297         nlh->nlmsg_len = skb->tail - b;
 298         return skb->len;
 299
 300 nlmsg_failure:
 301 nfattr_failure:
 302         skb_trim(skb, b - skb->data);
 303         return -1;
 304 }
 305
 306 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 307 static int ctnetlink_conntrack_event(struct notifier_block *this,
 308                                      unsigned long events, void *ptr)
 309 {
 310         struct nlmsghdr *nlh;
 311         struct nfgenmsg *nfmsg;
 312         struct nfattr *nest_parms;
 313         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 314         struct sk_buff *skb;
 315         unsigned int type;
 316         unsigned char *b;
 317         unsigned int flags = 0, group;
 318
 319         /* ignore our fake conntrack entry */
 320         if (ct == &ip_conntrack_untracked)
 321                 return NOTIFY_DONE;
 322
 323         if (events & IPCT_DESTROY) {
 324                 type = IPCTNL_MSG_CT_DELETE;
 325                 group = NFNLGRP_CONNTRACK_DESTROY;
 326         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 327                 type = IPCTNL_MSG_CT_NEW;
 328                 flags = NLM_F_CREATE|NLM_F_EXCL;
 329                 /* dump everything */
 330                 events = ~0UL;
 331                 group = NFNLGRP_CONNTRACK_NEW;
 332         } else if (events & (IPCT_STATUS |
 333                       IPCT_PROTOINFO |
 334                       IPCT_HELPER |
 335                       IPCT_HELPINFO |
 336                       IPCT_NATINFO)) {
 337                 type = IPCTNL_MSG_CT_NEW;
 338                 group = NFNLGRP_CONNTRACK_UPDATE;
 339         } else
 340                 return NOTIFY_DONE;
 341
 342         if (!nfnetlink_has_listeners(group))
 343                 return NOTIFY_DONE;
 344
 345         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 346         if (!skb)
 347                 return NOTIFY_DONE;
 348
 349         b = skb->tail;
 350
 351         type |= NFNL_SUBSYS_CTNETLINK << 8;
 352         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 353         nfmsg = NLMSG_DATA(nlh);
 354
 355         nlh->nlmsg_flags    = flags;
 356         nfmsg->nfgen_family = AF_INET;
 357         nfmsg->version  = NFNETLINK_V0;
 358         nfmsg->res_id   = 0;
 359
 360         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 361         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 362                 goto nfattr_failure;
 363         NFA_NEST_END(skb, nest_parms);
 364
 365         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 366         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 367                 goto nfattr_failure;
 368         NFA_NEST_END(skb, nest_parms);
 369
 370         /* NAT stuff is now a status flag */
 371         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 372             && ctnetlink_dump_status(skb, ct) < 0)
 373                 goto nfattr_failure;
 374         if (events & IPCT_REFRESH
 375             && ctnetlink_dump_timeout(skb, ct) < 0)
 376                 goto nfattr_failure;
 377         if (events & IPCT_PROTOINFO
 378             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 379                 goto nfattr_failure;
 380         if (events & IPCT_HELPINFO
 381             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 382                 goto nfattr_failure;
 383
 384         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 385             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 386                 goto nfattr_failure;
 387
 388         nlh->nlmsg_len = skb->tail - b;
 389         nfnetlink_send(skb, 0, group, 0);
 390         return NOTIFY_DONE;
 391
 392 nlmsg_failure:
 393 nfattr_failure:
 394         kfree_skb(skb);
 395         return NOTIFY_DONE;
 396 }
 397 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 398
 399 static int ctnetlink_done(struct netlink_callback *cb)
 400 {
 401         DEBUGP("entered %s\n", __FUNCTION__);
 402         if (cb->args[1])
 403                 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
 404         return 0;
 405 }
 406
 407 static int
 408 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 409 {
 410         struct ip_conntrack *ct, *last;
 411         struct ip_conntrack_tuple_hash *h;
 412         struct list_head *i;
 413
 414         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 415                         cb->args[0], *id);
 416
 417         read_lock_bh(&ip_conntrack_lock);
 418         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
 419 restart:
 420                 last = (struct ip_conntrack *)cb->args[1];
 421                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 422                         h = (struct ip_conntrack_tuple_hash *) i;
 423                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 424                                 continue;
 425                         ct = tuplehash_to_ctrack(h);
 426                         if (last != NULL) {
 427                                 if (ct == last) {
 428                                         ip_conntrack_put(last);
 429                                         cb->args[1] = 0;
 430                                         last = NULL;
 431                                 } else
 432                                         continue;
 433                         }
 434                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 435                                                 cb->nlh->nlmsg_seq,
 436                                                 IPCTNL_MSG_CT_NEW,
 437                                                 1, ct) < 0) {
 438                                 nf_conntrack_get(&ct->ct_general);
 439                                 cb->args[1] = (unsigned long)ct;
 440                                 goto out;
 441                         }
 442                 }
 443                 if (last != NULL) {
 444                         ip_conntrack_put(last);
 445                         cb->args[1] = 0;
 446                         goto restart;
 447                 }
 448         }
 449 out:
 450         read_unlock_bh(&ip_conntrack_lock);
 451
 452         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 453
 454         return skb->len;
 455 }
 456
 457 #ifdef CONFIG_IP_NF_CT_ACCT
 458 static int
 459 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 460 {
 461         struct ip_conntrack *ct = NULL;
 462         struct ip_conntrack_tuple_hash *h;
 463         struct list_head *i;
 464         u_int32_t *id = (u_int32_t *) &cb->args[1];
 465
 466         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 467                         cb->args[0], *id);
 468
 469         write_lock_bh(&ip_conntrack_lock);
 470         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 471                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 472                         h = (struct ip_conntrack_tuple_hash *) i;
 473                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 474                                 continue;
 475                         ct = tuplehash_to_ctrack(h);
 476                         if (ct->id <= *id)
 477                                 continue;
 478                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 479                                                 cb->nlh->nlmsg_seq,
 480                                                 IPCTNL_MSG_CT_NEW,
 481                                                 1, ct) < 0)
 482                                 goto out;
 483                         *id = ct->id;
 484
 485                         memset(&ct->counters, 0, sizeof(ct->counters));
 486                 }
 487         }
 488 out:
 489         write_unlock_bh(&ip_conntrack_lock);
 490
 491         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 492
 493         return skb->len;
 494 }
 495 #endif
 496
 497 static const size_t cta_min_ip[CTA_IP_MAX] = {
 498         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 499         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 500 };
 501
 502 static inline int
 503 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 504 {
 505         struct nfattr *tb[CTA_IP_MAX];
 506
 507         DEBUGP("entered %s\n", __FUNCTION__);
 508
 509         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 510
 511         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 512                 return -EINVAL;
 513
 514         if (!tb[CTA_IP_V4_SRC-1])
 515                 return -EINVAL;
 516         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 517
 518         if (!tb[CTA_IP_V4_DST-1])
 519                 return -EINVAL;
 520         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 521
 522         DEBUGP("leaving\n");
 523
 524         return 0;
 525 }
 526
 527 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 528         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 529         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 530         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 531         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 532         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 533         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 534 };
 535
 536 static inline int
 537 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 538                             struct ip_conntrack_tuple *tuple)
 539 {
 540         struct nfattr *tb[CTA_PROTO_MAX];
 541         struct ip_conntrack_protocol *proto;
 542         int ret = 0;
 543
 544         DEBUGP("entered %s\n", __FUNCTION__);
 545
 546         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 547
 548         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 549                 return -EINVAL;
 550
 551         if (!tb[CTA_PROTO_NUM-1])
 552                 return -EINVAL;
 553         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 554
 555         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 556
 557         if (likely(proto->nfattr_to_tuple))
 558                 ret = proto->nfattr_to_tuple(tb, tuple);
 559
 560         ip_conntrack_proto_put(proto);
 561
 562         return ret;
 563 }
 564
 565 static inline int
 566 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 567                       enum ctattr_tuple type)
 568 {
 569         struct nfattr *tb[CTA_TUPLE_MAX];
 570         int err;
 571
 572         DEBUGP("entered %s\n", __FUNCTION__);
 573
 574         memset(tuple, 0, sizeof(*tuple));
 575
 576         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 577
 578         if (!tb[CTA_TUPLE_IP-1])
 579                 return -EINVAL;
 580
 581         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 582         if (err < 0)
 583                 return err;
 584
 585         if (!tb[CTA_TUPLE_PROTO-1])
 586                 return -EINVAL;
 587
 588         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 589         if (err < 0)
 590                 return err;
 591
 592         /* orig and expect tuples get DIR_ORIGINAL */
 593         if (type == CTA_TUPLE_REPLY)
 594                 tuple->dst.dir = IP_CT_DIR_REPLY;
 595         else
 596                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 597
 598         DUMP_TUPLE(tuple);
 599
 600         DEBUGP("leaving\n");
 601
 602         return 0;
 603 }
 604
 605 #ifdef CONFIG_IP_NF_NAT_NEEDED
 606 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 607         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 608         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 609 };
 610
 611 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 612                                      const struct ip_conntrack *ct,
 613                                      struct ip_nat_range *range)
 614 {
 615         struct nfattr *tb[CTA_PROTONAT_MAX];
 616         struct ip_nat_protocol *npt;
 617
 618         DEBUGP("entered %s\n", __FUNCTION__);
 619
 620         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 621
 622         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 623                 return -EINVAL;
 624
 625         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 626
 627         if (!npt->nfattr_to_range) {
 628                 ip_nat_proto_put(npt);
 629                 return 0;
 630         }
 631
 632         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 633         if (npt->nfattr_to_range(tb, range) > 0)
 634                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 635
 636         ip_nat_proto_put(npt);
 637
 638         DEBUGP("leaving\n");
 639         return 0;
 640 }
 641
 642 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 643         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 644         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 645 };
 646
 647 static inline int
 648 ctnetlink_parse_nat(struct nfattr *nat,
 649                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 650 {
 651         struct nfattr *tb[CTA_NAT_MAX];
 652         int err;
 653
 654         DEBUGP("entered %s\n", __FUNCTION__);
 655
 656         memset(range, 0, sizeof(*range));
 657
 658         nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
 659
 660         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 661                 return -EINVAL;
 662
 663         if (tb[CTA_NAT_MINIP-1])
 664                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 665
 666         if (!tb[CTA_NAT_MAXIP-1])
 667                 range->max_ip = range->min_ip;
 668         else
 669                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 670
 671         if (range->min_ip)
 672                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 673
 674         if (!tb[CTA_NAT_PROTO-1])
 675                 return 0;
 676
 677         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 678         if (err < 0)
 679                 return err;
 680
 681         DEBUGP("leaving\n");
 682         return 0;
 683 }
 684 #endif
 685
 686 static inline int
 687 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 688 {
 689         struct nfattr *tb[CTA_HELP_MAX];
 690
 691         DEBUGP("entered %s\n", __FUNCTION__);
 692
 693         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 694
 695         if (!tb[CTA_HELP_NAME-1])
 696                 return -EINVAL;
 697
 698         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 699
 700         return 0;
 701 }
 702
 703 static const size_t cta_min[CTA_MAX] = {
 704         [CTA_STATUS-1]          = sizeof(u_int32_t),
 705         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 706         [CTA_MARK-1]            = sizeof(u_int32_t),
 707         [CTA_USE-1]             = sizeof(u_int32_t),
 708         [CTA_ID-1]              = sizeof(u_int32_t)
 709 };
 710
 711 static int
 712 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 713                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 714 {
 715         struct ip_conntrack_tuple_hash *h;
 716         struct ip_conntrack_tuple tuple;
 717         struct ip_conntrack *ct;
 718         int err = 0;
 719
 720         DEBUGP("entered %s\n", __FUNCTION__);
 721
 722         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 723                 return -EINVAL;
 724
 725         if (cda[CTA_TUPLE_ORIG-1])
 726                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 727         else if (cda[CTA_TUPLE_REPLY-1])
 728                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 729         else {
 730                 /* Flush the whole table */
 731                 ip_conntrack_flush();
 732                 return 0;
 733         }
 734
 735         if (err < 0)
 736                 return err;
 737
 738         h = ip_conntrack_find_get(&tuple, NULL);
 739         if (!h) {
 740                 DEBUGP("tuple not found in conntrack hash\n");
 741                 return -ENOENT;
 742         }
 743
 744         ct = tuplehash_to_ctrack(h);
 745
 746         if (cda[CTA_ID-1]) {
 747                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 748                 if (ct->id != id) {
 749                         ip_conntrack_put(ct);
 750                         return -ENOENT;
 751                 }
 752         }
 753         if (del_timer(&ct->timeout))
 754                 ct->timeout.function((unsigned long)ct);
 755
 756         ip_conntrack_put(ct);
 757         DEBUGP("leaving\n");
 758
 759         return 0;
 760 }
 761
 762 static int
 763 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 764                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 765 {
 766         struct ip_conntrack_tuple_hash *h;
 767         struct ip_conntrack_tuple tuple;
 768         struct ip_conntrack *ct;
 769         struct sk_buff *skb2 = NULL;
 770         int err = 0;
 771
 772         DEBUGP("entered %s\n", __FUNCTION__);
 773
 774         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 775                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 776                 u32 rlen;
 777
 778                 if (msg->nfgen_family != AF_INET)
 779                         return -EAFNOSUPPORT;
 780
 781                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 782                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 783 #ifdef CONFIG_IP_NF_CT_ACCT
 784                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 785                                                 ctnetlink_dump_table_w,
 786                                                 ctnetlink_done)) != 0)
 787                                 return -EINVAL;
 788 #else
 789                         return -ENOTSUPP;
 790 #endif
 791                 } else {
 792                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 793                                                         ctnetlink_dump_table,
 794                                                         ctnetlink_done)) != 0)
 795                         return -EINVAL;
 796                 }
 797
 798                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 799                 if (rlen > skb->len)
 800                         rlen = skb->len;
 801                 skb_pull(skb, rlen);
 802                 return 0;
 803         }
 804
 805         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 806                 return -EINVAL;
 807
 808         if (cda[CTA_TUPLE_ORIG-1])
 809                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 810         else if (cda[CTA_TUPLE_REPLY-1])
 811                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 812         else
 813                 return -EINVAL;
 814
 815         if (err < 0)
 816                 return err;
 817
 818         h = ip_conntrack_find_get(&tuple, NULL);
 819         if (!h) {
 820                 DEBUGP("tuple not found in conntrack hash");
 821                 return -ENOENT;
 822         }
 823         DEBUGP("tuple found\n");
 824         ct = tuplehash_to_ctrack(h);
 825
 826         err = -ENOMEM;
 827         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 828         if (!skb2) {
 829                 ip_conntrack_put(ct);
 830                 return -ENOMEM;
 831         }
 832         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 833
 834         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 835                                   IPCTNL_MSG_CT_NEW, 1, ct);
 836         ip_conntrack_put(ct);
 837         if (err <= 0)
 838                 goto free;
 839
 840         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 841         if (err < 0)
 842                 goto out;
 843
 844         DEBUGP("leaving\n");
 845         return 0;
 846
 847 free:
 848         kfree_skb(skb2);
 849 out:
 850         return err;
 851 }
 852
 853 static inline int
 854 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 855 {
 856         unsigned long d;
 857         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 858         d = ct->status ^ status;
 859
 860         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 861                 /* unchangeable */
 862                 return -EINVAL;
 863
 864         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 865                 /* SEEN_REPLY bit can only be set */
 866                 return -EINVAL;
 867
 868
 869         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 870                 /* ASSURED bit can only be set */
 871                 return -EINVAL;
 872
 873         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 874 #ifndef CONFIG_IP_NF_NAT_NEEDED
 875                 return -EINVAL;
 876 #else
 877                 struct ip_nat_range range;
 878
 879                 if (cda[CTA_NAT_DST-1]) {
 880                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 881                                                 &range) < 0)
 882                                 return -EINVAL;
 883                         if (ip_nat_initialized(ct,
 884                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 885                                 return -EEXIST;
 886                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 887                 }
 888                 if (cda[CTA_NAT_SRC-1]) {
 889                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 890                                                 &range) < 0)
 891                                 return -EINVAL;
 892                         if (ip_nat_initialized(ct,
 893                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 894                                 return -EEXIST;
 895                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 896                 }
 897 #endif
 898         }
 899
 900         /* Be careful here, modifying NAT bits can screw up things,
 901          * so don't let users modify them directly if they don't pass
 902          * ip_nat_range. */
 903         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 904         return 0;
 905 }
 906
 907
 908 static inline int
 909 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 910 {
 911         struct ip_conntrack_helper *helper;
 912         char *helpname;
 913         int err;
 914
 915         DEBUGP("entered %s\n", __FUNCTION__);
 916
 917         /* don't change helper of sibling connections */
 918         if (ct->master)
 919                 return -EINVAL;
 920
 921         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 922         if (err < 0)
 923                 return err;
 924
 925         helper = __ip_conntrack_helper_find_byname(helpname);
 926         if (!helper) {
 927                 if (!strcmp(helpname, ""))
 928                         helper = NULL;
 929                 else
 930                         return -EINVAL;
 931         }
 932
 933         if (ct->helper) {
 934                 if (!helper) {
 935                         /* we had a helper before ... */
 936                         ip_ct_remove_expectations(ct);
 937                         ct->helper = NULL;
 938                 } else {
 939                         /* need to zero data of old helper */
 940                         memset(&ct->help, 0, sizeof(ct->help));
 941                 }
 942         }
 943
 944         ct->helper = helper;
 945
 946         return 0;
 947 }
 948
 949 static inline int
 950 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 951 {
 952         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 953
 954         if (!del_timer(&ct->timeout))
 955                 return -ETIME;
 956
 957         ct->timeout.expires = jiffies + timeout * HZ;
 958         add_timer(&ct->timeout);
 959
 960         return 0;
 961 }
 962
 963 static inline int
 964 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 965 {
 966         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 967         struct ip_conntrack_protocol *proto;
 968         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 969         int err = 0;
 970
 971         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 972
 973         proto = ip_conntrack_proto_find_get(npt);
 974
 975         if (proto->from_nfattr)
 976                 err = proto->from_nfattr(tb, ct);
 977         ip_conntrack_proto_put(proto);
 978
 979         return err;
 980 }
 981
 982 static int
 983 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 984 {
 985         int err;
 986
 987         DEBUGP("entered %s\n", __FUNCTION__);
 988
 989         if (cda[CTA_HELP-1]) {
 990                 err = ctnetlink_change_helper(ct, cda);
 991                 if (err < 0)
 992                         return err;
 993         }
 994
 995         if (cda[CTA_TIMEOUT-1]) {
 996                 err = ctnetlink_change_timeout(ct, cda);
 997                 if (err < 0)
 998                         return err;
 999         }
1000
1001         if (cda[CTA_STATUS-1]) {
1002                 err = ctnetlink_change_status(ct, cda);
1003                 if (err < 0)
1004                         return err;
1005         }
1006
1007         if (cda[CTA_PROTOINFO-1]) {
1008                 err = ctnetlink_change_protoinfo(ct, cda);
1009                 if (err < 0)
1010                         return err;
1011         }
1012
1013 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1014         if (cda[CTA_MARK-1])
1015                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1016 #endif
1017
1018         DEBUGP("all done\n");
1019         return 0;
1020 }
1021
1022 static int
1023 ctnetlink_create_conntrack(struct nfattr *cda[],
1024                            struct ip_conntrack_tuple *otuple,
1025                            struct ip_conntrack_tuple *rtuple)
1026 {
1027         struct ip_conntrack *ct;
1028         int err = -EINVAL;
1029
1030         DEBUGP("entered %s\n", __FUNCTION__);
1031
1032         ct = ip_conntrack_alloc(otuple, rtuple);
1033         if (ct == NULL || IS_ERR(ct))
1034                 return -ENOMEM;
1035
1036         if (!cda[CTA_TIMEOUT-1])
1037                 goto err;
1038         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1039
1040         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1041         ct->status |= IPS_CONFIRMED;
1042
1043         err = ctnetlink_change_status(ct, cda);
1044         if (err < 0)
1045                 goto err;
1046
1047         if (cda[CTA_PROTOINFO-1]) {
1048                 err = ctnetlink_change_protoinfo(ct, cda);
1049                 if (err < 0)
1050                         return err;
1051         }
1052
1053 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1054         if (cda[CTA_MARK-1])
1055                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1056 #endif
1057
1058         ct->helper = ip_conntrack_helper_find_get(rtuple);
1059
1060         add_timer(&ct->timeout);
1061         ip_conntrack_hash_insert(ct);
1062
1063         if (ct->helper)
1064                 ip_conntrack_helper_put(ct->helper);
1065
1066         DEBUGP("conntrack with id %u inserted\n", ct->id);
1067         return 0;
1068
1069 err:
1070         ip_conntrack_free(ct);
1071         return err;
1072 }
1073
1074 static int
1075 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1076                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1077 {
1078         struct ip_conntrack_tuple otuple, rtuple;
1079         struct ip_conntrack_tuple_hash *h = NULL;
1080         int err = 0;
1081
1082         DEBUGP("entered %s\n", __FUNCTION__);
1083
1084         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1085                 return -EINVAL;
1086
1087         if (cda[CTA_TUPLE_ORIG-1]) {
1088                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1089                 if (err < 0)
1090                         return err;
1091         }
1092
1093         if (cda[CTA_TUPLE_REPLY-1]) {
1094                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1095                 if (err < 0)
1096                         return err;
1097         }
1098
1099         write_lock_bh(&ip_conntrack_lock);
1100         if (cda[CTA_TUPLE_ORIG-1])
1101                 h = __ip_conntrack_find(&otuple, NULL);
1102         else if (cda[CTA_TUPLE_REPLY-1])
1103                 h = __ip_conntrack_find(&rtuple, NULL);
1104
1105         if (h == NULL) {
1106                 write_unlock_bh(&ip_conntrack_lock);
1107                 DEBUGP("no such conntrack, create new\n");
1108                 err = -ENOENT;
1109                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1110                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1111                 return err;
1112         }
1113         /* implicit 'else' */
1114
1115         /* we only allow nat config for new conntracks */
1116         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1117                 err = -EINVAL;
1118                 goto out_unlock;
1119         }
1120
1121         /* We manipulate the conntrack inside the global conntrack table lock,
1122          * so there's no need to increase the refcount */
1123         DEBUGP("conntrack found\n");
1124         err = -EEXIST;
1125         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1126                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1127
1128 out_unlock:
1129         write_unlock_bh(&ip_conntrack_lock);
1130         return err;
1131 }
1132
1133 /***********************************************************************
1134  * EXPECT
1135  ***********************************************************************/
1136
1137 static inline int
1138 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1139                          const struct ip_conntrack_tuple *tuple,
1140                          enum ctattr_expect type)
1141 {
1142         struct nfattr *nest_parms = NFA_NEST(skb, type);
1143
1144         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1145                 goto nfattr_failure;
1146
1147         NFA_NEST_END(skb, nest_parms);
1148
1149         return 0;
1150
1151 nfattr_failure:
1152         return -1;
1153 }
1154
1155 static inline int
1156 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1157                         const struct ip_conntrack_tuple *tuple,
1158                         const struct ip_conntrack_tuple *mask)
1159 {
1160         int ret;
1161         struct ip_conntrack_protocol *proto;
1162         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1163
1164         ret = ctnetlink_dump_tuples_ip(skb, mask);
1165         if (unlikely(ret < 0))
1166                 goto nfattr_failure;
1167
1168         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1169         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1170         ip_conntrack_proto_put(proto);
1171         if (unlikely(ret < 0))
1172                 goto nfattr_failure;
1173
1174         NFA_NEST_END(skb, nest_parms);
1175
1176         return 0;
1177
1178 nfattr_failure:
1179         return -1;
1180 }
1181
1182 static inline int
1183 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1184                           const struct ip_conntrack_expect *exp)
1185 {
1186         struct ip_conntrack *master = exp->master;
1187         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1188         u_int32_t id = htonl(exp->id);
1189
1190         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1191                 goto nfattr_failure;
1192         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1193                 goto nfattr_failure;
1194         if (ctnetlink_exp_dump_tuple(skb,
1195                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1196                                  CTA_EXPECT_MASTER) < 0)
1197                 goto nfattr_failure;
1198
1199         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1200         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1201
1202         return 0;
1203
1204 nfattr_failure:
1205         return -1;
1206 }
1207
1208 static int
1209 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1210                     int event,
1211                     int nowait,
1212                     const struct ip_conntrack_expect *exp)
1213 {
1214         struct nlmsghdr *nlh;
1215         struct nfgenmsg *nfmsg;
1216         unsigned char *b;
1217
1218         b = skb->tail;
1219
1220         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1221         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1222         nfmsg  = NLMSG_DATA(nlh);
1223
1224         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1225         nfmsg->nfgen_family = AF_INET;
1226         nfmsg->version      = NFNETLINK_V0;
1227         nfmsg->res_id       = 0;
1228
1229         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1230                 goto nfattr_failure;
1231
1232         nlh->nlmsg_len = skb->tail - b;
1233         return skb->len;
1234
1235 nlmsg_failure:
1236 nfattr_failure:
1237         skb_trim(skb, b - skb->data);
1238         return -1;
1239 }
1240
1241 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1242 static int ctnetlink_expect_event(struct notifier_block *this,
1243                                   unsigned long events, void *ptr)
1244 {
1245         struct nlmsghdr *nlh;
1246         struct nfgenmsg *nfmsg;
1247         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1248         struct sk_buff *skb;
1249         unsigned int type;
1250         unsigned char *b;
1251         int flags = 0;
1252
1253         if (events & IPEXP_NEW) {
1254                 type = IPCTNL_MSG_EXP_NEW;
1255                 flags = NLM_F_CREATE|NLM_F_EXCL;
1256         } else
1257                 return NOTIFY_DONE;
1258
1259         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1260         if (!skb)
1261                 return NOTIFY_DONE;
1262
1263         b = skb->tail;
1264
1265         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1266         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1267         nfmsg = NLMSG_DATA(nlh);
1268
1269         nlh->nlmsg_flags    = flags;
1270         nfmsg->nfgen_family = AF_INET;
1271         nfmsg->version      = NFNETLINK_V0;
1272         nfmsg->res_id       = 0;
1273
1274         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1275                 goto nfattr_failure;
1276
1277         nlh->nlmsg_len = skb->tail - b;
1278         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1279         return NOTIFY_DONE;
1280
1281 nlmsg_failure:
1282 nfattr_failure:
1283         kfree_skb(skb);
1284         return NOTIFY_DONE;
1285 }
1286 #endif
1287
1288 static int
1289 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1290 {
1291         struct ip_conntrack_expect *exp = NULL;
1292         struct list_head *i;
1293         u_int32_t *id = (u_int32_t *) &cb->args[0];
1294
1295         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1296
1297         read_lock_bh(&ip_conntrack_lock);
1298         list_for_each_prev(i, &ip_conntrack_expect_list) {
1299                 exp = (struct ip_conntrack_expect *) i;
1300                 if (exp->id <= *id)
1301                         continue;
1302                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1303                                             cb->nlh->nlmsg_seq,
1304                                             IPCTNL_MSG_EXP_NEW,
1305                                             1, exp) < 0)
1306                         goto out;
1307                 *id = exp->id;
1308         }
1309 out:
1310         read_unlock_bh(&ip_conntrack_lock);
1311
1312         DEBUGP("leaving, last id=%llu\n", *id);
1313
1314         return skb->len;
1315 }
1316
1317 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1318         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1319         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1320 };
1321
1322 static int
1323 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1324                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1325 {
1326         struct ip_conntrack_tuple tuple;
1327         struct ip_conntrack_expect *exp;
1328         struct sk_buff *skb2;
1329         int err = 0;
1330
1331         DEBUGP("entered %s\n", __FUNCTION__);
1332
1333         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1334                 return -EINVAL;
1335
1336         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1337                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1338                 u32 rlen;
1339
1340                 if (msg->nfgen_family != AF_INET)
1341                         return -EAFNOSUPPORT;
1342
1343                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1344                                                 ctnetlink_exp_dump_table,
1345                                                 ctnetlink_done)) != 0)
1346                         return -EINVAL;
1347                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1348                 if (rlen > skb->len)
1349                         rlen = skb->len;
1350                 skb_pull(skb, rlen);
1351                 return 0;
1352         }
1353
1354         if (cda[CTA_EXPECT_MASTER-1])
1355                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1356         else
1357                 return -EINVAL;
1358
1359         if (err < 0)
1360                 return err;
1361
1362         exp = ip_conntrack_expect_find(&tuple);
1363         if (!exp)
1364                 return -ENOENT;
1365
1366         if (cda[CTA_EXPECT_ID-1]) {
1367                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1368                 if (exp->id != ntohl(id)) {
1369                         ip_conntrack_expect_put(exp);
1370                         return -ENOENT;
1371                 }
1372         }
1373
1374         err = -ENOMEM;
1375         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1376         if (!skb2)
1377                 goto out;
1378         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1379
1380         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1381                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1382                                       1, exp);
1383         if (err <= 0)
1384                 goto free;
1385
1386         ip_conntrack_expect_put(exp);
1387
1388         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1389
1390 free:
1391         kfree_skb(skb2);
1392 out:
1393         ip_conntrack_expect_put(exp);
1394         return err;
1395 }
1396
1397 static int
1398 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1399                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1400 {
1401         struct ip_conntrack_expect *exp, *tmp;
1402         struct ip_conntrack_tuple tuple;
1403         struct ip_conntrack_helper *h;
1404         int err;
1405
1406         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1407                 return -EINVAL;
1408
1409         if (cda[CTA_EXPECT_TUPLE-1]) {
1410                 /* delete a single expect by tuple */
1411                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1412                 if (err < 0)
1413                         return err;
1414
1415                 /* bump usage count to 2 */
1416                 exp = ip_conntrack_expect_find(&tuple);
1417                 if (!exp)
1418                         return -ENOENT;
1419
1420                 if (cda[CTA_EXPECT_ID-1]) {
1421                         u_int32_t id =
1422                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1423                         if (exp->id != ntohl(id)) {
1424                                 ip_conntrack_expect_put(exp);
1425                                 return -ENOENT;
1426                         }
1427                 }
1428
1429                 /* after list removal, usage count == 1 */
1430                 ip_conntrack_unexpect_related(exp);
1431                 /* have to put what we 'get' above.
1432                  * after this line usage count == 0 */
1433                 ip_conntrack_expect_put(exp);
1434         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1435                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1436
1437                 /* delete all expectations for this helper */
1438                 write_lock_bh(&ip_conntrack_lock);
1439                 h = __ip_conntrack_helper_find_byname(name);
1440                 if (!h) {
1441                         write_unlock_bh(&ip_conntrack_lock);
1442                         return -EINVAL;
1443                 }
1444                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1445                                          list) {
1446                         if (exp->master->helper == h
1447                             && del_timer(&exp->timeout)) {
1448                                 ip_ct_unlink_expect(exp);
1449                                 ip_conntrack_expect_put(exp);
1450                         }
1451                 }
1452                 write_unlock_bh(&ip_conntrack_lock);
1453         } else {
1454                 /* This basically means we have to flush everything*/
1455                 write_lock_bh(&ip_conntrack_lock);
1456                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1457                                          list) {
1458                         if (del_timer(&exp->timeout)) {
1459                                 ip_ct_unlink_expect(exp);
1460                                 ip_conntrack_expect_put(exp);
1461                         }
1462                 }
1463                 write_unlock_bh(&ip_conntrack_lock);
1464         }
1465
1466         return 0;
1467 }
1468 static int
1469 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1470 {
1471         return -EOPNOTSUPP;
1472 }
1473
1474 static int
1475 ctnetlink_create_expect(struct nfattr *cda[])
1476 {
1477         struct ip_conntrack_tuple tuple, mask, master_tuple;
1478         struct ip_conntrack_tuple_hash *h = NULL;
1479         struct ip_conntrack_expect *exp;
1480         struct ip_conntrack *ct;
1481         int err = 0;
1482
1483         DEBUGP("entered %s\n", __FUNCTION__);
1484
1485         /* caller guarantees that those three CTA_EXPECT_* exist */
1486         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1487         if (err < 0)
1488                 return err;
1489         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1490         if (err < 0)
1491                 return err;
1492         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1493         if (err < 0)
1494                 return err;
1495
1496         /* Look for master conntrack of this expectation */
1497         h = ip_conntrack_find_get(&master_tuple, NULL);
1498         if (!h)
1499                 return -ENOENT;
1500         ct = tuplehash_to_ctrack(h);
1501
1502         if (!ct->helper) {
1503                 /* such conntrack hasn't got any helper, abort */
1504                 err = -EINVAL;
1505                 goto out;
1506         }
1507
1508         exp = ip_conntrack_expect_alloc(ct);
1509         if (!exp) {
1510                 err = -ENOMEM;
1511                 goto out;
1512         }
1513
1514         exp->expectfn = NULL;
1515         exp->flags = 0;
1516         exp->master = ct;
1517         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1518         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1519
1520         err = ip_conntrack_expect_related(exp);
1521         ip_conntrack_expect_put(exp);
1522
1523 out:
1524         ip_conntrack_put(tuplehash_to_ctrack(h));
1525         return err;
1526 }
1527
1528 static int
1529 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1530                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1531 {
1532         struct ip_conntrack_tuple tuple;
1533         struct ip_conntrack_expect *exp;
1534         int err = 0;
1535
1536         DEBUGP("entered %s\n", __FUNCTION__);
1537
1538         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1539                 return -EINVAL;
1540
1541         if (!cda[CTA_EXPECT_TUPLE-1]
1542             || !cda[CTA_EXPECT_MASK-1]
1543             || !cda[CTA_EXPECT_MASTER-1])
1544                 return -EINVAL;
1545
1546         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1547         if (err < 0)
1548                 return err;
1549
1550         write_lock_bh(&ip_conntrack_lock);
1551         exp = __ip_conntrack_expect_find(&tuple);
1552
1553         if (!exp) {
1554                 write_unlock_bh(&ip_conntrack_lock);
1555                 err = -ENOENT;
1556                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1557                         err = ctnetlink_create_expect(cda);
1558                 return err;
1559         }
1560
1561         err = -EEXIST;
1562         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1563                 err = ctnetlink_change_expect(exp, cda);
1564         write_unlock_bh(&ip_conntrack_lock);
1565
1566         DEBUGP("leaving\n");
1567
1568         return err;
1569 }
1570
1571 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1572 static struct notifier_block ctnl_notifier = {
1573         .notifier_call  = ctnetlink_conntrack_event,
1574 };
1575
1576 static struct notifier_block ctnl_notifier_exp = {
1577         .notifier_call  = ctnetlink_expect_event,
1578 };
1579 #endif
1580
1581 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1582         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1583                                             .attr_count = CTA_MAX, },
1584         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1585                                             .attr_count = CTA_MAX, },
1586         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1587                                             .attr_count = CTA_MAX, },
1588         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1589                                             .attr_count = CTA_MAX, },
1590 };
1591
1592 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1593         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1594                                             .attr_count = CTA_EXPECT_MAX, },
1595         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1596                                             .attr_count = CTA_EXPECT_MAX, },
1597         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1598                                             .attr_count = CTA_EXPECT_MAX, },
1599 };
1600
1601 static struct nfnetlink_subsystem ctnl_subsys = {
1602         .name                           = "conntrack",
1603         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1604         .cb_count                       = IPCTNL_MSG_MAX,
1605         .cb                             = ctnl_cb,
1606 };
1607
1608 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1609         .name                           = "conntrack_expect",
1610         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1611         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1612         .cb                             = ctnl_exp_cb,
1613 };
1614
1615 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1616 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1617
1618 static int __init ctnetlink_init(void)
1619 {
1620         int ret;
1621
1622         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1623         ret = nfnetlink_subsys_register(&ctnl_subsys);
1624         if (ret < 0) {
1625                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1626                 goto err_out;
1627         }
1628
1629         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1630         if (ret < 0) {
1631                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1632                 goto err_unreg_subsys;
1633         }
1634
1635 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1636         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1637         if (ret < 0) {
1638                 printk("ctnetlink_init: cannot register notifier.\n");
1639                 goto err_unreg_exp_subsys;
1640         }
1641
1642         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1643         if (ret < 0) {
1644                 printk("ctnetlink_init: cannot expect register notifier.\n");
1645                 goto err_unreg_notifier;
1646         }
1647 #endif
1648
1649         return 0;
1650
1651 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1652 err_unreg_notifier:
1653         ip_conntrack_unregister_notifier(&ctnl_notifier);
1654 err_unreg_exp_subsys:
1655         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1656 #endif
1657 err_unreg_subsys:
1658         nfnetlink_subsys_unregister(&ctnl_subsys);
1659 err_out:
1660         return ret;
1661 }
1662
1663 static void __exit ctnetlink_exit(void)
1664 {
1665         printk("ctnetlink: unregistering from nfnetlink.\n");
1666
1667 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1668         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1669         ip_conntrack_unregister_notifier(&ctnl_notifier);
1670 #endif
1671
1672         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1673         nfnetlink_subsys_unregister(&ctnl_subsys);
1674         return;
1675 }
1676
1677 module_init(ctnetlink_init);
1678 module_exit(ctnetlink_exit);