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