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         if (events & IPCT_MARK
 389             && ctnetlink_dump_mark(skb, ct) < 0)
 390                 goto nfattr_failure;
 391
 392         nlh->nlmsg_len = skb->tail - b;
 393         nfnetlink_send(skb, 0, group, 0);
 394         return NOTIFY_DONE;
 395
 396 nlmsg_failure:
 397 nfattr_failure:
 398         kfree_skb(skb);
 399         return NOTIFY_DONE;
 400 }
 401 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 402
 403 static int ctnetlink_done(struct netlink_callback *cb)
 404 {
 405         DEBUGP("entered %s\n", __FUNCTION__);
 406         if (cb->args[1])
 407                 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
 408         return 0;
 409 }
 410
 411 static int
 412 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 413 {
 414         struct ip_conntrack *ct, *last;
 415         struct ip_conntrack_tuple_hash *h;
 416         struct list_head *i;
 417
 418         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 419                         cb->args[0], *id);
 420
 421         read_lock_bh(&ip_conntrack_lock);
 422         last = (struct ip_conntrack *)cb->args[1];
 423         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
 424 restart:
 425                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 426                         h = (struct ip_conntrack_tuple_hash *) i;
 427                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 428                                 continue;
 429                         ct = tuplehash_to_ctrack(h);
 430                         if (cb->args[1]) {
 431                                 if (ct != last)
 432                                         continue;
 433                                 cb->args[1] = 0;
 434                         }
 435                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 436                                                 cb->nlh->nlmsg_seq,
 437                                                 IPCTNL_MSG_CT_NEW,
 438                                                 1, ct) < 0) {
 439                                 nf_conntrack_get(&ct->ct_general);
 440                                 cb->args[1] = (unsigned long)ct;
 441                                 goto out;
 442                         }
 443                 }
 444                 if (cb->args[1]) {
 445                         cb->args[1] = 0;
 446                         goto restart;
 447                 }
 448         }
 449 out:
 450         read_unlock_bh(&ip_conntrack_lock);
 451         if (last)
 452                 ip_conntrack_put(last);
 453
 454         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 455         return skb->len;
 456 }
 457
 458 #ifdef CONFIG_IP_NF_CT_ACCT
 459 static int
 460 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 461 {
 462         struct ip_conntrack *ct = NULL;
 463         struct ip_conntrack_tuple_hash *h;
 464         struct list_head *i;
 465         u_int32_t *id = (u_int32_t *) &cb->args[1];
 466
 467         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 468                         cb->args[0], *id);
 469
 470         write_lock_bh(&ip_conntrack_lock);
 471         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 472                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 473                         h = (struct ip_conntrack_tuple_hash *) i;
 474                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 475                                 continue;
 476                         ct = tuplehash_to_ctrack(h);
 477                         if (ct->id <= *id)
 478                                 continue;
 479                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 480                                                 cb->nlh->nlmsg_seq,
 481                                                 IPCTNL_MSG_CT_NEW,
 482                                                 1, ct) < 0)
 483                                 goto out;
 484                         *id = ct->id;
 485
 486                         memset(&ct->counters, 0, sizeof(ct->counters));
 487                 }
 488         }
 489 out:
 490         write_unlock_bh(&ip_conntrack_lock);
 491
 492         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 493
 494         return skb->len;
 495 }
 496 #endif
 497
 498 static const size_t cta_min_ip[CTA_IP_MAX] = {
 499         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 500         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 501 };
 502
 503 static inline int
 504 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 505 {
 506         struct nfattr *tb[CTA_IP_MAX];
 507
 508         DEBUGP("entered %s\n", __FUNCTION__);
 509
 510         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 511
 512         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 513                 return -EINVAL;
 514
 515         if (!tb[CTA_IP_V4_SRC-1])
 516                 return -EINVAL;
 517         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 518
 519         if (!tb[CTA_IP_V4_DST-1])
 520                 return -EINVAL;
 521         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 522
 523         DEBUGP("leaving\n");
 524
 525         return 0;
 526 }
 527
 528 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 529         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 530         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 531         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 532         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 533         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 534         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 535 };
 536
 537 static inline int
 538 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 539                             struct ip_conntrack_tuple *tuple)
 540 {
 541         struct nfattr *tb[CTA_PROTO_MAX];
 542         struct ip_conntrack_protocol *proto;
 543         int ret = 0;
 544
 545         DEBUGP("entered %s\n", __FUNCTION__);
 546
 547         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 548
 549         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 550                 return -EINVAL;
 551
 552         if (!tb[CTA_PROTO_NUM-1])
 553                 return -EINVAL;
 554         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 555
 556         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 557
 558         if (likely(proto->nfattr_to_tuple))
 559                 ret = proto->nfattr_to_tuple(tb, tuple);
 560
 561         ip_conntrack_proto_put(proto);
 562
 563         return ret;
 564 }
 565
 566 static inline int
 567 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 568                       enum ctattr_tuple type)
 569 {
 570         struct nfattr *tb[CTA_TUPLE_MAX];
 571         int err;
 572
 573         DEBUGP("entered %s\n", __FUNCTION__);
 574
 575         memset(tuple, 0, sizeof(*tuple));
 576
 577         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 578
 579         if (!tb[CTA_TUPLE_IP-1])
 580                 return -EINVAL;
 581
 582         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 583         if (err < 0)
 584                 return err;
 585
 586         if (!tb[CTA_TUPLE_PROTO-1])
 587                 return -EINVAL;
 588
 589         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 590         if (err < 0)
 591                 return err;
 592
 593         /* orig and expect tuples get DIR_ORIGINAL */
 594         if (type == CTA_TUPLE_REPLY)
 595                 tuple->dst.dir = IP_CT_DIR_REPLY;
 596         else
 597                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 598
 599         DUMP_TUPLE(tuple);
 600
 601         DEBUGP("leaving\n");
 602
 603         return 0;
 604 }
 605
 606 #ifdef CONFIG_IP_NF_NAT_NEEDED
 607 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 608         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 609         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 610 };
 611
 612 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 613                                      const struct ip_conntrack *ct,
 614                                      struct ip_nat_range *range)
 615 {
 616         struct nfattr *tb[CTA_PROTONAT_MAX];
 617         struct ip_nat_protocol *npt;
 618
 619         DEBUGP("entered %s\n", __FUNCTION__);
 620
 621         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 622
 623         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 624                 return -EINVAL;
 625
 626         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 627
 628         if (!npt->nfattr_to_range) {
 629                 ip_nat_proto_put(npt);
 630                 return 0;
 631         }
 632
 633         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 634         if (npt->nfattr_to_range(tb, range) > 0)
 635                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 636
 637         ip_nat_proto_put(npt);
 638
 639         DEBUGP("leaving\n");
 640         return 0;
 641 }
 642
 643 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 644         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 645         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 646 };
 647
 648 static inline int
 649 ctnetlink_parse_nat(struct nfattr *nat,
 650                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 651 {
 652         struct nfattr *tb[CTA_NAT_MAX];
 653         int err;
 654
 655         DEBUGP("entered %s\n", __FUNCTION__);
 656
 657         memset(range, 0, sizeof(*range));
 658
 659         nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
 660
 661         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 662                 return -EINVAL;
 663
 664         if (tb[CTA_NAT_MINIP-1])
 665                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 666
 667         if (!tb[CTA_NAT_MAXIP-1])
 668                 range->max_ip = range->min_ip;
 669         else
 670                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 671
 672         if (range->min_ip)
 673                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 674
 675         if (!tb[CTA_NAT_PROTO-1])
 676                 return 0;
 677
 678         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 679         if (err < 0)
 680                 return err;
 681
 682         DEBUGP("leaving\n");
 683         return 0;
 684 }
 685 #endif
 686
 687 static inline int
 688 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 689 {
 690         struct nfattr *tb[CTA_HELP_MAX];
 691
 692         DEBUGP("entered %s\n", __FUNCTION__);
 693
 694         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 695
 696         if (!tb[CTA_HELP_NAME-1])
 697                 return -EINVAL;
 698
 699         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 700
 701         return 0;
 702 }
 703
 704 static const size_t cta_min[CTA_MAX] = {
 705         [CTA_STATUS-1]          = sizeof(u_int32_t),
 706         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 707         [CTA_MARK-1]            = sizeof(u_int32_t),
 708         [CTA_USE-1]             = sizeof(u_int32_t),
 709         [CTA_ID-1]              = sizeof(u_int32_t)
 710 };
 711
 712 static int
 713 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 714                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 715 {
 716         struct ip_conntrack_tuple_hash *h;
 717         struct ip_conntrack_tuple tuple;
 718         struct ip_conntrack *ct;
 719         int err = 0;
 720
 721         DEBUGP("entered %s\n", __FUNCTION__);
 722
 723         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 724                 return -EINVAL;
 725
 726         if (cda[CTA_TUPLE_ORIG-1])
 727                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 728         else if (cda[CTA_TUPLE_REPLY-1])
 729                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 730         else {
 731                 /* Flush the whole table */
 732                 ip_conntrack_flush();
 733                 return 0;
 734         }
 735
 736         if (err < 0)
 737                 return err;
 738
 739         h = ip_conntrack_find_get(&tuple, NULL);
 740         if (!h) {
 741                 DEBUGP("tuple not found in conntrack hash\n");
 742                 return -ENOENT;
 743         }
 744
 745         ct = tuplehash_to_ctrack(h);
 746
 747         if (cda[CTA_ID-1]) {
 748                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 749                 if (ct->id != id) {
 750                         ip_conntrack_put(ct);
 751                         return -ENOENT;
 752                 }
 753         }
 754         if (del_timer(&ct->timeout))
 755                 ct->timeout.function((unsigned long)ct);
 756
 757         ip_conntrack_put(ct);
 758         DEBUGP("leaving\n");
 759
 760         return 0;
 761 }
 762
 763 static int
 764 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 765                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 766 {
 767         struct ip_conntrack_tuple_hash *h;
 768         struct ip_conntrack_tuple tuple;
 769         struct ip_conntrack *ct;
 770         struct sk_buff *skb2 = NULL;
 771         int err = 0;
 772
 773         DEBUGP("entered %s\n", __FUNCTION__);
 774
 775         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 776                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 777                 u32 rlen;
 778
 779                 if (msg->nfgen_family != AF_INET)
 780                         return -EAFNOSUPPORT;
 781
 782                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 783                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 784 #ifdef CONFIG_IP_NF_CT_ACCT
 785                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 786                                                 ctnetlink_dump_table_w,
 787                                                 ctnetlink_done)) != 0)
 788                                 return -EINVAL;
 789 #else
 790                         return -ENOTSUPP;
 791 #endif
 792                 } else {
 793                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 794                                                         ctnetlink_dump_table,
 795                                                         ctnetlink_done)) != 0)
 796                         return -EINVAL;
 797                 }
 798
 799                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 800                 if (rlen > skb->len)
 801                         rlen = skb->len;
 802                 skb_pull(skb, rlen);
 803                 return 0;
 804         }
 805
 806         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 807                 return -EINVAL;
 808
 809         if (cda[CTA_TUPLE_ORIG-1])
 810                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 811         else if (cda[CTA_TUPLE_REPLY-1])
 812                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 813         else
 814                 return -EINVAL;
 815
 816         if (err < 0)
 817                 return err;
 818
 819         h = ip_conntrack_find_get(&tuple, NULL);
 820         if (!h) {
 821                 DEBUGP("tuple not found in conntrack hash");
 822                 return -ENOENT;
 823         }
 824         DEBUGP("tuple found\n");
 825         ct = tuplehash_to_ctrack(h);
 826
 827         err = -ENOMEM;
 828         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 829         if (!skb2) {
 830                 ip_conntrack_put(ct);
 831                 return -ENOMEM;
 832         }
 833         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 834
 835         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 836                                   IPCTNL_MSG_CT_NEW, 1, ct);
 837         ip_conntrack_put(ct);
 838         if (err <= 0)
 839                 goto free;
 840
 841         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 842         if (err < 0)
 843                 goto out;
 844
 845         DEBUGP("leaving\n");
 846         return 0;
 847
 848 free:
 849         kfree_skb(skb2);
 850 out:
 851         return err;
 852 }
 853
 854 static inline int
 855 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 856 {
 857         unsigned long d;
 858         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 859         d = ct->status ^ status;
 860
 861         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 862                 /* unchangeable */
 863                 return -EINVAL;
 864
 865         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 866                 /* SEEN_REPLY bit can only be set */
 867                 return -EINVAL;
 868
 869
 870         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 871                 /* ASSURED bit can only be set */
 872                 return -EINVAL;
 873
 874         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 875 #ifndef CONFIG_IP_NF_NAT_NEEDED
 876                 return -EINVAL;
 877 #else
 878                 struct ip_nat_range range;
 879
 880                 if (cda[CTA_NAT_DST-1]) {
 881                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 882                                                 &range) < 0)
 883                                 return -EINVAL;
 884                         if (ip_nat_initialized(ct,
 885                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 886                                 return -EEXIST;
 887                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 888                 }
 889                 if (cda[CTA_NAT_SRC-1]) {
 890                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 891                                                 &range) < 0)
 892                                 return -EINVAL;
 893                         if (ip_nat_initialized(ct,
 894                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 895                                 return -EEXIST;
 896                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 897                 }
 898 #endif
 899         }
 900
 901         /* Be careful here, modifying NAT bits can screw up things,
 902          * so don't let users modify them directly if they don't pass
 903          * ip_nat_range. */
 904         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 905         return 0;
 906 }
 907
 908
 909 static inline int
 910 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 911 {
 912         struct ip_conntrack_helper *helper;
 913         char *helpname;
 914         int err;
 915
 916         DEBUGP("entered %s\n", __FUNCTION__);
 917
 918         /* don't change helper of sibling connections */
 919         if (ct->master)
 920                 return -EINVAL;
 921
 922         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 923         if (err < 0)
 924                 return err;
 925
 926         helper = __ip_conntrack_helper_find_byname(helpname);
 927         if (!helper) {
 928                 if (!strcmp(helpname, ""))
 929                         helper = NULL;
 930                 else
 931                         return -EINVAL;
 932         }
 933
 934         if (ct->helper) {
 935                 if (!helper) {
 936                         /* we had a helper before ... */
 937                         ip_ct_remove_expectations(ct);
 938                         ct->helper = NULL;
 939                 } else {
 940                         /* need to zero data of old helper */
 941                         memset(&ct->help, 0, sizeof(ct->help));
 942                 }
 943         }
 944
 945         ct->helper = helper;
 946
 947         return 0;
 948 }
 949
 950 static inline int
 951 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 952 {
 953         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 954
 955         if (!del_timer(&ct->timeout))
 956                 return -ETIME;
 957
 958         ct->timeout.expires = jiffies + timeout * HZ;
 959         add_timer(&ct->timeout);
 960
 961         return 0;
 962 }
 963
 964 static inline int
 965 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 966 {
 967         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 968         struct ip_conntrack_protocol *proto;
 969         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 970         int err = 0;
 971
 972         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 973
 974         proto = ip_conntrack_proto_find_get(npt);
 975
 976         if (proto->from_nfattr)
 977                 err = proto->from_nfattr(tb, ct);
 978         ip_conntrack_proto_put(proto);
 979
 980         return err;
 981 }
 982
 983 static int
 984 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 985 {
 986         int err;
 987
 988         DEBUGP("entered %s\n", __FUNCTION__);
 989
 990         if (cda[CTA_HELP-1]) {
 991                 err = ctnetlink_change_helper(ct, cda);
 992                 if (err < 0)
 993                         return err;
 994         }
 995
 996         if (cda[CTA_TIMEOUT-1]) {
 997                 err = ctnetlink_change_timeout(ct, cda);
 998                 if (err < 0)
 999                         return err;
1000         }
1001
1002         if (cda[CTA_STATUS-1]) {
1003                 err = ctnetlink_change_status(ct, cda);
1004                 if (err < 0)
1005                         return err;
1006         }
1007
1008         if (cda[CTA_PROTOINFO-1]) {
1009                 err = ctnetlink_change_protoinfo(ct, cda);
1010                 if (err < 0)
1011                         return err;
1012         }
1013
1014 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1015         if (cda[CTA_MARK-1])
1016                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1017 #endif
1018
1019         DEBUGP("all done\n");
1020         return 0;
1021 }
1022
1023 static int
1024 ctnetlink_create_conntrack(struct nfattr *cda[],
1025                            struct ip_conntrack_tuple *otuple,
1026                            struct ip_conntrack_tuple *rtuple)
1027 {
1028         struct ip_conntrack *ct;
1029         int err = -EINVAL;
1030
1031         DEBUGP("entered %s\n", __FUNCTION__);
1032
1033         ct = ip_conntrack_alloc(otuple, rtuple);
1034         if (ct == NULL || IS_ERR(ct))
1035                 return -ENOMEM;
1036
1037         if (!cda[CTA_TIMEOUT-1])
1038                 goto err;
1039         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1040
1041         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1042         ct->status |= IPS_CONFIRMED;
1043
1044         err = ctnetlink_change_status(ct, cda);
1045         if (err < 0)
1046                 goto err;
1047
1048         if (cda[CTA_PROTOINFO-1]) {
1049                 err = ctnetlink_change_protoinfo(ct, cda);
1050                 if (err < 0)
1051                         return err;
1052         }
1053
1054 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1055         if (cda[CTA_MARK-1])
1056                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1057 #endif
1058
1059         ct->helper = ip_conntrack_helper_find_get(rtuple);
1060
1061         add_timer(&ct->timeout);
1062         ip_conntrack_hash_insert(ct);
1063
1064         if (ct->helper)
1065                 ip_conntrack_helper_put(ct->helper);
1066
1067         DEBUGP("conntrack with id %u inserted\n", ct->id);
1068         return 0;
1069
1070 err:
1071         ip_conntrack_free(ct);
1072         return err;
1073 }
1074
1075 static int
1076 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1077                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1078 {
1079         struct ip_conntrack_tuple otuple, rtuple;
1080         struct ip_conntrack_tuple_hash *h = NULL;
1081         int err = 0;
1082
1083         DEBUGP("entered %s\n", __FUNCTION__);
1084
1085         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1086                 return -EINVAL;
1087
1088         if (cda[CTA_TUPLE_ORIG-1]) {
1089                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1090                 if (err < 0)
1091                         return err;
1092         }
1093
1094         if (cda[CTA_TUPLE_REPLY-1]) {
1095                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1096                 if (err < 0)
1097                         return err;
1098         }
1099
1100         write_lock_bh(&ip_conntrack_lock);
1101         if (cda[CTA_TUPLE_ORIG-1])
1102                 h = __ip_conntrack_find(&otuple, NULL);
1103         else if (cda[CTA_TUPLE_REPLY-1])
1104                 h = __ip_conntrack_find(&rtuple, NULL);
1105
1106         if (h == NULL) {
1107                 write_unlock_bh(&ip_conntrack_lock);
1108                 DEBUGP("no such conntrack, create new\n");
1109                 err = -ENOENT;
1110                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1111                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1112                 return err;
1113         }
1114         /* implicit 'else' */
1115
1116         /* we only allow nat config for new conntracks */
1117         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1118                 err = -EINVAL;
1119                 goto out_unlock;
1120         }
1121
1122         /* We manipulate the conntrack inside the global conntrack table lock,
1123          * so there's no need to increase the refcount */
1124         DEBUGP("conntrack found\n");
1125         err = -EEXIST;
1126         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1127                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1128
1129 out_unlock:
1130         write_unlock_bh(&ip_conntrack_lock);
1131         return err;
1132 }
1133
1134 /***********************************************************************
1135  * EXPECT
1136  ***********************************************************************/
1137
1138 static inline int
1139 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1140                          const struct ip_conntrack_tuple *tuple,
1141                          enum ctattr_expect type)
1142 {
1143         struct nfattr *nest_parms = NFA_NEST(skb, type);
1144
1145         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1146                 goto nfattr_failure;
1147
1148         NFA_NEST_END(skb, nest_parms);
1149
1150         return 0;
1151
1152 nfattr_failure:
1153         return -1;
1154 }
1155
1156 static inline int
1157 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1158                         const struct ip_conntrack_tuple *tuple,
1159                         const struct ip_conntrack_tuple *mask)
1160 {
1161         int ret;
1162         struct ip_conntrack_protocol *proto;
1163         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1164
1165         ret = ctnetlink_dump_tuples_ip(skb, mask);
1166         if (unlikely(ret < 0))
1167                 goto nfattr_failure;
1168
1169         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1170         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1171         ip_conntrack_proto_put(proto);
1172         if (unlikely(ret < 0))
1173                 goto nfattr_failure;
1174
1175         NFA_NEST_END(skb, nest_parms);
1176
1177         return 0;
1178
1179 nfattr_failure:
1180         return -1;
1181 }
1182
1183 static inline int
1184 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1185                           const struct ip_conntrack_expect *exp)
1186 {
1187         struct ip_conntrack *master = exp->master;
1188         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1189         u_int32_t id = htonl(exp->id);
1190
1191         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1192                 goto nfattr_failure;
1193         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1194                 goto nfattr_failure;
1195         if (ctnetlink_exp_dump_tuple(skb,
1196                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1197                                  CTA_EXPECT_MASTER) < 0)
1198                 goto nfattr_failure;
1199
1200         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1201         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1202
1203         return 0;
1204
1205 nfattr_failure:
1206         return -1;
1207 }
1208
1209 static int
1210 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1211                     int event,
1212                     int nowait,
1213                     const struct ip_conntrack_expect *exp)
1214 {
1215         struct nlmsghdr *nlh;
1216         struct nfgenmsg *nfmsg;
1217         unsigned char *b;
1218
1219         b = skb->tail;
1220
1221         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1222         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1223         nfmsg  = NLMSG_DATA(nlh);
1224
1225         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1226         nfmsg->nfgen_family = AF_INET;
1227         nfmsg->version      = NFNETLINK_V0;
1228         nfmsg->res_id       = 0;
1229
1230         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1231                 goto nfattr_failure;
1232
1233         nlh->nlmsg_len = skb->tail - b;
1234         return skb->len;
1235
1236 nlmsg_failure:
1237 nfattr_failure:
1238         skb_trim(skb, b - skb->data);
1239         return -1;
1240 }
1241
1242 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1243 static int ctnetlink_expect_event(struct notifier_block *this,
1244                                   unsigned long events, void *ptr)
1245 {
1246         struct nlmsghdr *nlh;
1247         struct nfgenmsg *nfmsg;
1248         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1249         struct sk_buff *skb;
1250         unsigned int type;
1251         unsigned char *b;
1252         int flags = 0;
1253
1254         if (events & IPEXP_NEW) {
1255                 type = IPCTNL_MSG_EXP_NEW;
1256                 flags = NLM_F_CREATE|NLM_F_EXCL;
1257         } else
1258                 return NOTIFY_DONE;
1259
1260         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1261                 return NOTIFY_DONE;
1262
1263         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1264         if (!skb)
1265                 return NOTIFY_DONE;
1266
1267         b = skb->tail;
1268
1269         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1270         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1271         nfmsg = NLMSG_DATA(nlh);
1272
1273         nlh->nlmsg_flags    = flags;
1274         nfmsg->nfgen_family = AF_INET;
1275         nfmsg->version      = NFNETLINK_V0;
1276         nfmsg->res_id       = 0;
1277
1278         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1279                 goto nfattr_failure;
1280
1281         nlh->nlmsg_len = skb->tail - b;
1282         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1283         return NOTIFY_DONE;
1284
1285 nlmsg_failure:
1286 nfattr_failure:
1287         kfree_skb(skb);
1288         return NOTIFY_DONE;
1289 }
1290 #endif
1291
1292 static int
1293 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1294 {
1295         struct ip_conntrack_expect *exp = NULL;
1296         struct list_head *i;
1297         u_int32_t *id = (u_int32_t *) &cb->args[0];
1298
1299         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1300
1301         read_lock_bh(&ip_conntrack_lock);
1302         list_for_each_prev(i, &ip_conntrack_expect_list) {
1303                 exp = (struct ip_conntrack_expect *) i;
1304                 if (exp->id <= *id)
1305                         continue;
1306                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1307                                             cb->nlh->nlmsg_seq,
1308                                             IPCTNL_MSG_EXP_NEW,
1309                                             1, exp) < 0)
1310                         goto out;
1311                 *id = exp->id;
1312         }
1313 out:
1314         read_unlock_bh(&ip_conntrack_lock);
1315
1316         DEBUGP("leaving, last id=%llu\n", *id);
1317
1318         return skb->len;
1319 }
1320
1321 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1322         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1323         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1324 };
1325
1326 static int
1327 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1328                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1329 {
1330         struct ip_conntrack_tuple tuple;
1331         struct ip_conntrack_expect *exp;
1332         struct sk_buff *skb2;
1333         int err = 0;
1334
1335         DEBUGP("entered %s\n", __FUNCTION__);
1336
1337         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1338                 return -EINVAL;
1339
1340         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1341                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1342                 u32 rlen;
1343
1344                 if (msg->nfgen_family != AF_INET)
1345                         return -EAFNOSUPPORT;
1346
1347                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1348                                                 ctnetlink_exp_dump_table,
1349                                                 ctnetlink_done)) != 0)
1350                         return -EINVAL;
1351                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1352                 if (rlen > skb->len)
1353                         rlen = skb->len;
1354                 skb_pull(skb, rlen);
1355                 return 0;
1356         }
1357
1358         if (cda[CTA_EXPECT_MASTER-1])
1359                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1360         else
1361                 return -EINVAL;
1362
1363         if (err < 0)
1364                 return err;
1365
1366         exp = ip_conntrack_expect_find(&tuple);
1367         if (!exp)
1368                 return -ENOENT;
1369
1370         if (cda[CTA_EXPECT_ID-1]) {
1371                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1372                 if (exp->id != ntohl(id)) {
1373                         ip_conntrack_expect_put(exp);
1374                         return -ENOENT;
1375                 }
1376         }
1377
1378         err = -ENOMEM;
1379         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1380         if (!skb2)
1381                 goto out;
1382         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1383
1384         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1385                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1386                                       1, exp);
1387         if (err <= 0)
1388                 goto free;
1389
1390         ip_conntrack_expect_put(exp);
1391
1392         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1393
1394 free:
1395         kfree_skb(skb2);
1396 out:
1397         ip_conntrack_expect_put(exp);
1398         return err;
1399 }
1400
1401 static int
1402 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1403                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1404 {
1405         struct ip_conntrack_expect *exp, *tmp;
1406         struct ip_conntrack_tuple tuple;
1407         struct ip_conntrack_helper *h;
1408         int err;
1409
1410         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1411                 return -EINVAL;
1412
1413         if (cda[CTA_EXPECT_TUPLE-1]) {
1414                 /* delete a single expect by tuple */
1415                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1416                 if (err < 0)
1417                         return err;
1418
1419                 /* bump usage count to 2 */
1420                 exp = ip_conntrack_expect_find(&tuple);
1421                 if (!exp)
1422                         return -ENOENT;
1423
1424                 if (cda[CTA_EXPECT_ID-1]) {
1425                         u_int32_t id =
1426                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1427                         if (exp->id != ntohl(id)) {
1428                                 ip_conntrack_expect_put(exp);
1429                                 return -ENOENT;
1430                         }
1431                 }
1432
1433                 /* after list removal, usage count == 1 */
1434                 ip_conntrack_unexpect_related(exp);
1435                 /* have to put what we 'get' above.
1436                  * after this line usage count == 0 */
1437                 ip_conntrack_expect_put(exp);
1438         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1439                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1440
1441                 /* delete all expectations for this helper */
1442                 write_lock_bh(&ip_conntrack_lock);
1443                 h = __ip_conntrack_helper_find_byname(name);
1444                 if (!h) {
1445                         write_unlock_bh(&ip_conntrack_lock);
1446                         return -EINVAL;
1447                 }
1448                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1449                                          list) {
1450                         if (exp->master->helper == h
1451                             && del_timer(&exp->timeout)) {
1452                                 ip_ct_unlink_expect(exp);
1453                                 ip_conntrack_expect_put(exp);
1454                         }
1455                 }
1456                 write_unlock_bh(&ip_conntrack_lock);
1457         } else {
1458                 /* This basically means we have to flush everything*/
1459                 write_lock_bh(&ip_conntrack_lock);
1460                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1461                                          list) {
1462                         if (del_timer(&exp->timeout)) {
1463                                 ip_ct_unlink_expect(exp);
1464                                 ip_conntrack_expect_put(exp);
1465                         }
1466                 }
1467                 write_unlock_bh(&ip_conntrack_lock);
1468         }
1469
1470         return 0;
1471 }
1472 static int
1473 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1474 {
1475         return -EOPNOTSUPP;
1476 }
1477
1478 static int
1479 ctnetlink_create_expect(struct nfattr *cda[])
1480 {
1481         struct ip_conntrack_tuple tuple, mask, master_tuple;
1482         struct ip_conntrack_tuple_hash *h = NULL;
1483         struct ip_conntrack_expect *exp;
1484         struct ip_conntrack *ct;
1485         int err = 0;
1486
1487         DEBUGP("entered %s\n", __FUNCTION__);
1488
1489         /* caller guarantees that those three CTA_EXPECT_* exist */
1490         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1491         if (err < 0)
1492                 return err;
1493         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1494         if (err < 0)
1495                 return err;
1496         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1497         if (err < 0)
1498                 return err;
1499
1500         /* Look for master conntrack of this expectation */
1501         h = ip_conntrack_find_get(&master_tuple, NULL);
1502         if (!h)
1503                 return -ENOENT;
1504         ct = tuplehash_to_ctrack(h);
1505
1506         if (!ct->helper) {
1507                 /* such conntrack hasn't got any helper, abort */
1508                 err = -EINVAL;
1509                 goto out;
1510         }
1511
1512         exp = ip_conntrack_expect_alloc(ct);
1513         if (!exp) {
1514                 err = -ENOMEM;
1515                 goto out;
1516         }
1517
1518         exp->expectfn = NULL;
1519         exp->flags = 0;
1520         exp->master = ct;
1521         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1522         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1523
1524         err = ip_conntrack_expect_related(exp);
1525         ip_conntrack_expect_put(exp);
1526
1527 out:
1528         ip_conntrack_put(tuplehash_to_ctrack(h));
1529         return err;
1530 }
1531
1532 static int
1533 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1534                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1535 {
1536         struct ip_conntrack_tuple tuple;
1537         struct ip_conntrack_expect *exp;
1538         int err = 0;
1539
1540         DEBUGP("entered %s\n", __FUNCTION__);
1541
1542         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1543                 return -EINVAL;
1544
1545         if (!cda[CTA_EXPECT_TUPLE-1]
1546             || !cda[CTA_EXPECT_MASK-1]
1547             || !cda[CTA_EXPECT_MASTER-1])
1548                 return -EINVAL;
1549
1550         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1551         if (err < 0)
1552                 return err;
1553
1554         write_lock_bh(&ip_conntrack_lock);
1555         exp = __ip_conntrack_expect_find(&tuple);
1556
1557         if (!exp) {
1558                 write_unlock_bh(&ip_conntrack_lock);
1559                 err = -ENOENT;
1560                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1561                         err = ctnetlink_create_expect(cda);
1562                 return err;
1563         }
1564
1565         err = -EEXIST;
1566         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1567                 err = ctnetlink_change_expect(exp, cda);
1568         write_unlock_bh(&ip_conntrack_lock);
1569
1570         DEBUGP("leaving\n");
1571
1572         return err;
1573 }
1574
1575 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1576 static struct notifier_block ctnl_notifier = {
1577         .notifier_call  = ctnetlink_conntrack_event,
1578 };
1579
1580 static struct notifier_block ctnl_notifier_exp = {
1581         .notifier_call  = ctnetlink_expect_event,
1582 };
1583 #endif
1584
1585 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1586         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1587                                             .attr_count = CTA_MAX, },
1588         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1589                                             .attr_count = CTA_MAX, },
1590         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1591                                             .attr_count = CTA_MAX, },
1592         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1593                                             .attr_count = CTA_MAX, },
1594 };
1595
1596 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1597         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1598                                             .attr_count = CTA_EXPECT_MAX, },
1599         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1600                                             .attr_count = CTA_EXPECT_MAX, },
1601         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1602                                             .attr_count = CTA_EXPECT_MAX, },
1603 };
1604
1605 static struct nfnetlink_subsystem ctnl_subsys = {
1606         .name                           = "conntrack",
1607         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1608         .cb_count                       = IPCTNL_MSG_MAX,
1609         .cb                             = ctnl_cb,
1610 };
1611
1612 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1613         .name                           = "conntrack_expect",
1614         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1615         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1616         .cb                             = ctnl_exp_cb,
1617 };
1618
1619 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1620 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1621
1622 static int __init ctnetlink_init(void)
1623 {
1624         int ret;
1625
1626         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1627         ret = nfnetlink_subsys_register(&ctnl_subsys);
1628         if (ret < 0) {
1629                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1630                 goto err_out;
1631         }
1632
1633         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1634         if (ret < 0) {
1635                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1636                 goto err_unreg_subsys;
1637         }
1638
1639 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1640         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1641         if (ret < 0) {
1642                 printk("ctnetlink_init: cannot register notifier.\n");
1643                 goto err_unreg_exp_subsys;
1644         }
1645
1646         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1647         if (ret < 0) {
1648                 printk("ctnetlink_init: cannot expect register notifier.\n");
1649                 goto err_unreg_notifier;
1650         }
1651 #endif
1652
1653         return 0;
1654
1655 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1656 err_unreg_notifier:
1657         ip_conntrack_unregister_notifier(&ctnl_notifier);
1658 err_unreg_exp_subsys:
1659         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1660 #endif
1661 err_unreg_subsys:
1662         nfnetlink_subsys_unregister(&ctnl_subsys);
1663 err_out:
1664         return ret;
1665 }
1666
1667 static void __exit ctnetlink_exit(void)
1668 {
1669         printk("ctnetlink: unregistering from nfnetlink.\n");
1670
1671 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1672         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1673         ip_conntrack_unregister_notifier(&ctnl_notifier);
1674 #endif
1675
1676         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1677         nfnetlink_subsys_unregister(&ctnl_subsys);
1678         return;
1679 }
1680
1681 module_init(ctnetlink_init);
1682 module_exit(ctnetlink_exit);