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

   1 /* Connection tracking via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
   7  * (C) 2005 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 {
  58         struct ip_conntrack_protocol *proto;
  59         int ret = 0;
  60
  61         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  62
  63         /* If no protocol helper is found, this function will return the
  64          * generic protocol helper, so proto won't *ever* be NULL */
  65         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  66         if (likely(proto->tuple_to_nfattr))
  67                 ret = proto->tuple_to_nfattr(skb, tuple);
  68
  69         ip_conntrack_proto_put(proto);
  70
  71         return ret;
  72
  73 nfattr_failure:
  74         return -1;
  75 }
  76
  77 static inline int
  78 ctnetlink_dump_tuples(struct sk_buff *skb,
  79                       const struct ip_conntrack_tuple *tuple)
  80 {
  81         struct nfattr *nest_parms;
  82
  83         nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  84         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  85         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  86         NFA_NEST_END(skb, nest_parms);
  87
  88         nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  89         ctnetlink_dump_tuples_proto(skb, tuple);
  90         NFA_NEST_END(skb, nest_parms);
  91
  92         return 0;
  93
  94 nfattr_failure:
  95         return -1;
  96 }
  97
  98 static inline int
  99 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 100 {
 101         u_int32_t status = htonl((u_int32_t) ct->status);
 102         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 103         return 0;
 104
 105 nfattr_failure:
 106         return -1;
 107 }
 108
 109 static inline int
 110 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 111 {
 112         long timeout_l = ct->timeout.expires - jiffies;
 113         u_int32_t timeout;
 114
 115         if (timeout_l < 0)
 116                 timeout = 0;
 117         else
 118                 timeout = htonl(timeout_l / HZ);
 119
 120         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 121         return 0;
 122
 123 nfattr_failure:
 124         return -1;
 125 }
 126
 127 static inline int
 128 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 129 {
 130         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 131
 132         struct nfattr *nest_proto;
 133         int ret;
 134
 135         if (!proto->to_nfattr) {
 136                 ip_conntrack_proto_put(proto);
 137                 return 0;
 138         }
 139
 140         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 141
 142         ret = proto->to_nfattr(skb, nest_proto, ct);
 143
 144         ip_conntrack_proto_put(proto);
 145
 146         NFA_NEST_END(skb, nest_proto);
 147
 148         return ret;
 149
 150 nfattr_failure:
 151         return -1;
 152 }
 153
 154 static inline int
 155 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 156 {
 157         struct nfattr *nest_helper;
 158
 159         if (!ct->helper)
 160                 return 0;
 161
 162         nest_helper = NFA_NEST(skb, CTA_HELP);
 163         NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
 164
 165         if (ct->helper->to_nfattr)
 166                 ct->helper->to_nfattr(skb, ct);
 167
 168         NFA_NEST_END(skb, nest_helper);
 169
 170         return 0;
 171
 172 nfattr_failure:
 173         return -1;
 174 }
 175
 176 #ifdef CONFIG_IP_NF_CT_ACCT
 177 static inline int
 178 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 179                         enum ip_conntrack_dir dir)
 180 {
 181         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 182         struct nfattr *nest_count = NFA_NEST(skb, type);
 183         u_int32_t tmp;
 184
 185         tmp = htonl(ct->counters[dir].packets);
 186         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 187
 188         tmp = htonl(ct->counters[dir].bytes);
 189         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 190
 191         NFA_NEST_END(skb, nest_count);
 192
 193         return 0;
 194
 195 nfattr_failure:
 196         return -1;
 197 }
 198 #else
 199 #define ctnetlink_dump_counters(a, b, c) (0)
 200 #endif
 201
 202 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 203 static inline int
 204 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 205 {
 206         u_int32_t mark = htonl(ct->mark);
 207
 208         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 209         return 0;
 210
 211 nfattr_failure:
 212         return -1;
 213 }
 214 #else
 215 #define ctnetlink_dump_mark(a, b) (0)
 216 #endif
 217
 218 static inline int
 219 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 220 {
 221         u_int32_t id = htonl(ct->id);
 222         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 223         return 0;
 224
 225 nfattr_failure:
 226         return -1;
 227 }
 228
 229 static inline int
 230 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 231 {
 232         unsigned int use = htonl(atomic_read(&ct->ct_general.use));
 233
 234         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 235         return 0;
 236
 237 nfattr_failure:
 238         return -1;
 239 }
 240
 241 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 242
 243 static int
 244 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 245                     int event, int nowait,
 246                     const struct ip_conntrack *ct)
 247 {
 248         struct nlmsghdr *nlh;
 249         struct nfgenmsg *nfmsg;
 250         struct nfattr *nest_parms;
 251         unsigned char *b;
 252
 253         b = skb->tail;
 254
 255         event |= NFNL_SUBSYS_CTNETLINK << 8;
 256         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 257         nfmsg  = NLMSG_DATA(nlh);
 258
 259         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 260         nfmsg->nfgen_family = AF_INET;
 261         nfmsg->version      = NFNETLINK_V0;
 262         nfmsg->res_id       = 0;
 263
 264         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 265         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 266                 goto nfattr_failure;
 267         NFA_NEST_END(skb, nest_parms);
 268
 269         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 270         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 271                 goto nfattr_failure;
 272         NFA_NEST_END(skb, nest_parms);
 273
 274         if (ctnetlink_dump_status(skb, ct) < 0 ||
 275             ctnetlink_dump_timeout(skb, ct) < 0 ||
 276             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 277             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 278             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 279             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 280             ctnetlink_dump_mark(skb, ct) < 0 ||
 281             ctnetlink_dump_id(skb, ct) < 0 ||
 282             ctnetlink_dump_use(skb, ct) < 0)
 283                 goto nfattr_failure;
 284
 285         nlh->nlmsg_len = skb->tail - b;
 286         return skb->len;
 287
 288 nlmsg_failure:
 289 nfattr_failure:
 290         skb_trim(skb, b - skb->data);
 291         return -1;
 292 }
 293
 294 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 295 static int ctnetlink_conntrack_event(struct notifier_block *this,
 296                                      unsigned long events, void *ptr)
 297 {
 298         struct nlmsghdr *nlh;
 299         struct nfgenmsg *nfmsg;
 300         struct nfattr *nest_parms;
 301         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 302         struct sk_buff *skb;
 303         unsigned int type;
 304         unsigned char *b;
 305         unsigned int flags = 0, group;
 306
 307         /* ignore our fake conntrack entry */
 308         if (ct == &ip_conntrack_untracked)
 309                 return NOTIFY_DONE;
 310
 311         if (events & IPCT_DESTROY) {
 312                 type = IPCTNL_MSG_CT_DELETE;
 313                 group = NFNLGRP_CONNTRACK_DESTROY;
 314                 goto alloc_skb;
 315         }
 316         if (events & (IPCT_NEW | IPCT_RELATED)) {
 317                 type = IPCTNL_MSG_CT_NEW;
 318                 flags = NLM_F_CREATE|NLM_F_EXCL;
 319                 /* dump everything */
 320                 events = ~0UL;
 321                 group = NFNLGRP_CONNTRACK_NEW;
 322                 goto alloc_skb;
 323         }
 324         if (events & (IPCT_STATUS |
 325                       IPCT_PROTOINFO |
 326                       IPCT_HELPER |
 327                       IPCT_HELPINFO |
 328                       IPCT_NATINFO)) {
 329                 type = IPCTNL_MSG_CT_NEW;
 330                 group = NFNLGRP_CONNTRACK_UPDATE;
 331                 goto alloc_skb;
 332         }
 333
 334         return NOTIFY_DONE;
 335
 336 alloc_skb:
 337   /* FIXME: Check if there are any listeners before, don't hurt performance */
 338
 339         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 340         if (!skb)
 341                 return NOTIFY_DONE;
 342
 343         b = skb->tail;
 344
 345         type |= NFNL_SUBSYS_CTNETLINK << 8;
 346         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 347         nfmsg = NLMSG_DATA(nlh);
 348
 349         nlh->nlmsg_flags    = flags;
 350         nfmsg->nfgen_family = AF_INET;
 351         nfmsg->version  = NFNETLINK_V0;
 352         nfmsg->res_id   = 0;
 353
 354         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 355         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 356                 goto nfattr_failure;
 357         NFA_NEST_END(skb, nest_parms);
 358
 359         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 360         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 361                 goto nfattr_failure;
 362         NFA_NEST_END(skb, nest_parms);
 363
 364         /* NAT stuff is now a status flag */
 365         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 366             && ctnetlink_dump_status(skb, ct) < 0)
 367                 goto nfattr_failure;
 368         if (events & IPCT_REFRESH
 369             && ctnetlink_dump_timeout(skb, ct) < 0)
 370                 goto nfattr_failure;
 371         if (events & IPCT_PROTOINFO
 372             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 373                 goto nfattr_failure;
 374         if (events & IPCT_HELPINFO
 375             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 376                 goto nfattr_failure;
 377
 378         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 379             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 380                 goto nfattr_failure;
 381
 382         nlh->nlmsg_len = skb->tail - b;
 383         nfnetlink_send(skb, 0, group, 0);
 384         return NOTIFY_DONE;
 385
 386 nlmsg_failure:
 387 nfattr_failure:
 388         kfree_skb(skb);
 389         return NOTIFY_DONE;
 390 }
 391 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 392
 393 static int ctnetlink_done(struct netlink_callback *cb)
 394 {
 395         DEBUGP("entered %s\n", __FUNCTION__);
 396         return 0;
 397 }
 398
 399 static int
 400 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 401 {
 402         struct ip_conntrack *ct = NULL;
 403         struct ip_conntrack_tuple_hash *h;
 404         struct list_head *i;
 405         u_int32_t *id = (u_int32_t *) &cb->args[1];
 406
 407         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 408                         cb->args[0], *id);
 409
 410         read_lock_bh(&ip_conntrack_lock);
 411         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 412                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 413                         h = (struct ip_conntrack_tuple_hash *) i;
 414                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 415                                 continue;
 416                         ct = tuplehash_to_ctrack(h);
 417                         if (ct->id <= *id)
 418                                 continue;
 419                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 420                                                 cb->nlh->nlmsg_seq,
 421                                                 IPCTNL_MSG_CT_NEW,
 422                                                 1, ct) < 0)
 423                                 goto out;
 424                         *id = ct->id;
 425                 }
 426         }
 427 out:
 428         read_unlock_bh(&ip_conntrack_lock);
 429
 430         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 431
 432         return skb->len;
 433 }
 434
 435 #ifdef CONFIG_IP_NF_CT_ACCT
 436 static int
 437 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 438 {
 439         struct ip_conntrack *ct = NULL;
 440         struct ip_conntrack_tuple_hash *h;
 441         struct list_head *i;
 442         u_int32_t *id = (u_int32_t *) &cb->args[1];
 443
 444         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 445                         cb->args[0], *id);
 446
 447         write_lock_bh(&ip_conntrack_lock);
 448         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 449                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 450                         h = (struct ip_conntrack_tuple_hash *) i;
 451                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 452                                 continue;
 453                         ct = tuplehash_to_ctrack(h);
 454                         if (ct->id <= *id)
 455                                 continue;
 456                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 457                                                 cb->nlh->nlmsg_seq,
 458                                                 IPCTNL_MSG_CT_NEW,
 459                                                 1, ct) < 0)
 460                                 goto out;
 461                         *id = ct->id;
 462
 463                         memset(&ct->counters, 0, sizeof(ct->counters));
 464                 }
 465         }
 466 out:
 467         write_unlock_bh(&ip_conntrack_lock);
 468
 469         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 470
 471         return skb->len;
 472 }
 473 #endif
 474
 475 static const size_t cta_min_ip[CTA_IP_MAX] = {
 476         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 477         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 478 };
 479
 480 static inline int
 481 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 482 {
 483         struct nfattr *tb[CTA_IP_MAX];
 484
 485         DEBUGP("entered %s\n", __FUNCTION__);
 486
 487         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 488
 489         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 490                 return -EINVAL;
 491
 492         if (!tb[CTA_IP_V4_SRC-1])
 493                 return -EINVAL;
 494         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 495
 496         if (!tb[CTA_IP_V4_DST-1])
 497                 return -EINVAL;
 498         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 499
 500         DEBUGP("leaving\n");
 501
 502         return 0;
 503 }
 504
 505 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 506         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 507         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 508         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 509         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 510         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 511         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 512 };
 513
 514 static inline int
 515 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 516                             struct ip_conntrack_tuple *tuple)
 517 {
 518         struct nfattr *tb[CTA_PROTO_MAX];
 519         struct ip_conntrack_protocol *proto;
 520         int ret = 0;
 521
 522         DEBUGP("entered %s\n", __FUNCTION__);
 523
 524         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 525
 526         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 527                 return -EINVAL;
 528
 529         if (!tb[CTA_PROTO_NUM-1])
 530                 return -EINVAL;
 531         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 532
 533         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 534
 535         if (likely(proto->nfattr_to_tuple))
 536                 ret = proto->nfattr_to_tuple(tb, tuple);
 537
 538         ip_conntrack_proto_put(proto);
 539
 540         return ret;
 541 }
 542
 543 static inline int
 544 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 545                       enum ctattr_tuple type)
 546 {
 547         struct nfattr *tb[CTA_TUPLE_MAX];
 548         int err;
 549
 550         DEBUGP("entered %s\n", __FUNCTION__);
 551
 552         memset(tuple, 0, sizeof(*tuple));
 553
 554         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 555
 556         if (!tb[CTA_TUPLE_IP-1])
 557                 return -EINVAL;
 558
 559         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 560         if (err < 0)
 561                 return err;
 562
 563         if (!tb[CTA_TUPLE_PROTO-1])
 564                 return -EINVAL;
 565
 566         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 567         if (err < 0)
 568                 return err;
 569
 570         /* orig and expect tuples get DIR_ORIGINAL */
 571         if (type == CTA_TUPLE_REPLY)
 572                 tuple->dst.dir = IP_CT_DIR_REPLY;
 573         else
 574                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 575
 576         DUMP_TUPLE(tuple);
 577
 578         DEBUGP("leaving\n");
 579
 580         return 0;
 581 }
 582
 583 #ifdef CONFIG_IP_NF_NAT_NEEDED
 584 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 585         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 586         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 587 };
 588
 589 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 590                                      const struct ip_conntrack *ct,
 591                                      struct ip_nat_range *range)
 592 {
 593         struct nfattr *tb[CTA_PROTONAT_MAX];
 594         struct ip_nat_protocol *npt;
 595
 596         DEBUGP("entered %s\n", __FUNCTION__);
 597
 598         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 599
 600         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 601                 return -EINVAL;
 602
 603         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 604
 605         if (!npt->nfattr_to_range) {
 606                 ip_nat_proto_put(npt);
 607                 return 0;
 608         }
 609
 610         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 611         if (npt->nfattr_to_range(tb, range) > 0)
 612                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 613
 614         ip_nat_proto_put(npt);
 615
 616         DEBUGP("leaving\n");
 617         return 0;
 618 }
 619
 620 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 621         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 622         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 623 };
 624
 625 static inline int
 626 ctnetlink_parse_nat(struct nfattr *cda[],
 627                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 628 {
 629         struct nfattr *tb[CTA_NAT_MAX];
 630         int err;
 631
 632         DEBUGP("entered %s\n", __FUNCTION__);
 633
 634         memset(range, 0, sizeof(*range));
 635
 636         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 637
 638         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 639                 return -EINVAL;
 640
 641         if (tb[CTA_NAT_MINIP-1])
 642                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 643
 644         if (!tb[CTA_NAT_MAXIP-1])
 645                 range->max_ip = range->min_ip;
 646         else
 647                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 648
 649         if (range->min_ip)
 650                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 651
 652         if (!tb[CTA_NAT_PROTO-1])
 653                 return 0;
 654
 655         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 656         if (err < 0)
 657                 return err;
 658
 659         DEBUGP("leaving\n");
 660         return 0;
 661 }
 662 #endif
 663
 664 static inline int
 665 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 666 {
 667         struct nfattr *tb[CTA_HELP_MAX];
 668
 669         DEBUGP("entered %s\n", __FUNCTION__);
 670
 671         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 672
 673         if (!tb[CTA_HELP_NAME-1])
 674                 return -EINVAL;
 675
 676         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 677
 678         return 0;
 679 }
 680
 681 static const size_t cta_min[CTA_MAX] = {
 682         [CTA_STATUS-1]          = sizeof(u_int32_t),
 683         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 684         [CTA_MARK-1]            = sizeof(u_int32_t),
 685         [CTA_USE-1]             = sizeof(u_int32_t),
 686         [CTA_ID-1]              = sizeof(u_int32_t)
 687 };
 688
 689 static int
 690 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 691                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 692 {
 693         struct ip_conntrack_tuple_hash *h;
 694         struct ip_conntrack_tuple tuple;
 695         struct ip_conntrack *ct;
 696         int err = 0;
 697
 698         DEBUGP("entered %s\n", __FUNCTION__);
 699
 700         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 701                 return -EINVAL;
 702
 703         if (cda[CTA_TUPLE_ORIG-1])
 704                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 705         else if (cda[CTA_TUPLE_REPLY-1])
 706                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 707         else {
 708                 /* Flush the whole table */
 709                 ip_conntrack_flush();
 710                 return 0;
 711         }
 712
 713         if (err < 0)
 714                 return err;
 715
 716         h = ip_conntrack_find_get(&tuple, NULL);
 717         if (!h) {
 718                 DEBUGP("tuple not found in conntrack hash\n");
 719                 return -ENOENT;
 720         }
 721
 722         ct = tuplehash_to_ctrack(h);
 723
 724         if (cda[CTA_ID-1]) {
 725                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 726                 if (ct->id != id) {
 727                         ip_conntrack_put(ct);
 728                         return -ENOENT;
 729                 }
 730         }
 731         if (del_timer(&ct->timeout))
 732                 ct->timeout.function((unsigned long)ct);
 733
 734         ip_conntrack_put(ct);
 735         DEBUGP("leaving\n");
 736
 737         return 0;
 738 }
 739
 740 static int
 741 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 742                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 743 {
 744         struct ip_conntrack_tuple_hash *h;
 745         struct ip_conntrack_tuple tuple;
 746         struct ip_conntrack *ct;
 747         struct sk_buff *skb2 = NULL;
 748         int err = 0;
 749
 750         DEBUGP("entered %s\n", __FUNCTION__);
 751
 752         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 753                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 754                 u32 rlen;
 755
 756                 if (msg->nfgen_family != AF_INET)
 757                         return -EAFNOSUPPORT;
 758
 759                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 760                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 761 #ifdef CONFIG_IP_NF_CT_ACCT
 762                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 763                                                 ctnetlink_dump_table_w,
 764                                                 ctnetlink_done)) != 0)
 765                                 return -EINVAL;
 766 #else
 767                         return -ENOTSUPP;
 768 #endif
 769                 } else {
 770                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 771                                                         ctnetlink_dump_table,
 772                                                         ctnetlink_done)) != 0)
 773                         return -EINVAL;
 774                 }
 775
 776                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 777                 if (rlen > skb->len)
 778                         rlen = skb->len;
 779                 skb_pull(skb, rlen);
 780                 return 0;
 781         }
 782
 783         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 784                 return -EINVAL;
 785
 786         if (cda[CTA_TUPLE_ORIG-1])
 787                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 788         else if (cda[CTA_TUPLE_REPLY-1])
 789                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 790         else
 791                 return -EINVAL;
 792
 793         if (err < 0)
 794                 return err;
 795
 796         h = ip_conntrack_find_get(&tuple, NULL);
 797         if (!h) {
 798                 DEBUGP("tuple not found in conntrack hash");
 799                 return -ENOENT;
 800         }
 801         DEBUGP("tuple found\n");
 802         ct = tuplehash_to_ctrack(h);
 803
 804         err = -ENOMEM;
 805         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 806         if (!skb2) {
 807                 ip_conntrack_put(ct);
 808                 return -ENOMEM;
 809         }
 810         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 811
 812         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 813                                   IPCTNL_MSG_CT_NEW, 1, ct);
 814         ip_conntrack_put(ct);
 815         if (err <= 0)
 816                 goto free;
 817
 818         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 819         if (err < 0)
 820                 goto out;
 821
 822         DEBUGP("leaving\n");
 823         return 0;
 824
 825 free:
 826         kfree_skb(skb2);
 827 out:
 828         return err;
 829 }
 830
 831 static inline int
 832 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 833 {
 834         unsigned long d;
 835         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 836         d = ct->status ^ status;
 837
 838         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 839                 /* unchangeable */
 840                 return -EINVAL;
 841
 842         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 843                 /* SEEN_REPLY bit can only be set */
 844                 return -EINVAL;
 845
 846
 847         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 848                 /* ASSURED bit can only be set */
 849                 return -EINVAL;
 850
 851         if (cda[CTA_NAT-1]) {
 852 #ifndef CONFIG_IP_NF_NAT_NEEDED
 853                 return -EINVAL;
 854 #else
 855                 unsigned int hooknum;
 856                 struct ip_nat_range range;
 857
 858                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 859                         return -EINVAL;
 860
 861                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 862                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 863                        htons(range.min.all), htons(range.max.all));
 864
 865                 /* This is tricky but it works. ip_nat_setup_info needs the
 866                  * hook number as parameter, so let's do the correct
 867                  * conversion and run away */
 868                 if (status & IPS_SRC_NAT_DONE)
 869                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 870                 else if (status & IPS_DST_NAT_DONE)
 871                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 872                 else
 873                         return -EINVAL; /* Missing NAT flags */
 874
 875                 DEBUGP("NAT status: %lu\n",
 876                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 877
 878                 if (ip_nat_initialized(ct, HOOK2MANIP(hooknum)))
 879                         return -EEXIST;
 880                 ip_nat_setup_info(ct, &range, hooknum);
 881
 882                 DEBUGP("NAT status after setup_info: %lu\n",
 883                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 884 #endif
 885         }
 886
 887         /* Be careful here, modifying NAT bits can screw up things,
 888          * so don't let users modify them directly if they don't pass
 889          * ip_nat_range. */
 890         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 891         return 0;
 892 }
 893
 894
 895 static inline int
 896 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 897 {
 898         struct ip_conntrack_helper *helper;
 899         char *helpname;
 900         int err;
 901
 902         DEBUGP("entered %s\n", __FUNCTION__);
 903
 904         /* don't change helper of sibling connections */
 905         if (ct->master)
 906                 return -EINVAL;
 907
 908         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 909         if (err < 0)
 910                 return err;
 911
 912         helper = __ip_conntrack_helper_find_byname(helpname);
 913         if (!helper) {
 914                 if (!strcmp(helpname, ""))
 915                         helper = NULL;
 916                 else
 917                         return -EINVAL;
 918         }
 919
 920         if (ct->helper) {
 921                 if (!helper) {
 922                         /* we had a helper before ... */
 923                         ip_ct_remove_expectations(ct);
 924                         ct->helper = NULL;
 925                 } else {
 926                         /* need to zero data of old helper */
 927                         memset(&ct->help, 0, sizeof(ct->help));
 928                 }
 929         }
 930
 931         ct->helper = helper;
 932
 933         return 0;
 934 }
 935
 936 static inline int
 937 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 938 {
 939         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 940
 941         if (!del_timer(&ct->timeout))
 942                 return -ETIME;
 943
 944         ct->timeout.expires = jiffies + timeout * HZ;
 945         add_timer(&ct->timeout);
 946
 947         return 0;
 948 }
 949
 950 static inline int
 951 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 952 {
 953         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 954         struct ip_conntrack_protocol *proto;
 955         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 956         int err = 0;
 957
 958         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 959
 960         proto = ip_conntrack_proto_find_get(npt);
 961
 962         if (proto->from_nfattr)
 963                 err = proto->from_nfattr(tb, ct);
 964         ip_conntrack_proto_put(proto);
 965
 966         return err;
 967 }
 968
 969 static int
 970 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 971 {
 972         int err;
 973
 974         DEBUGP("entered %s\n", __FUNCTION__);
 975
 976         if (cda[CTA_HELP-1]) {
 977                 err = ctnetlink_change_helper(ct, cda);
 978                 if (err < 0)
 979                         return err;
 980         }
 981
 982         if (cda[CTA_TIMEOUT-1]) {
 983                 err = ctnetlink_change_timeout(ct, cda);
 984                 if (err < 0)
 985                         return err;
 986         }
 987
 988         if (cda[CTA_STATUS-1]) {
 989                 err = ctnetlink_change_status(ct, cda);
 990                 if (err < 0)
 991                         return err;
 992         }
 993
 994         if (cda[CTA_PROTOINFO-1]) {
 995                 err = ctnetlink_change_protoinfo(ct, cda);
 996                 if (err < 0)
 997                         return err;
 998         }
 999
1000 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1001         if (cda[CTA_MARK-1])
1002                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1003 #endif
1004
1005         DEBUGP("all done\n");
1006         return 0;
1007 }
1008
1009 static int
1010 ctnetlink_create_conntrack(struct nfattr *cda[],
1011                            struct ip_conntrack_tuple *otuple,
1012                            struct ip_conntrack_tuple *rtuple)
1013 {
1014         struct ip_conntrack *ct;
1015         int err = -EINVAL;
1016
1017         DEBUGP("entered %s\n", __FUNCTION__);
1018
1019         ct = ip_conntrack_alloc(otuple, rtuple);
1020         if (ct == NULL || IS_ERR(ct))
1021                 return -ENOMEM;
1022
1023         if (!cda[CTA_TIMEOUT-1])
1024                 goto err;
1025         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1026
1027         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1028         ct->status |= IPS_CONFIRMED;
1029
1030         err = ctnetlink_change_status(ct, cda);
1031         if (err < 0)
1032                 goto err;
1033
1034         if (cda[CTA_PROTOINFO-1]) {
1035                 err = ctnetlink_change_protoinfo(ct, cda);
1036                 if (err < 0)
1037                         return err;
1038         }
1039
1040         ct->helper = ip_conntrack_helper_find_get(rtuple);
1041
1042         add_timer(&ct->timeout);
1043         ip_conntrack_hash_insert(ct);
1044
1045         if (ct->helper)
1046                 ip_conntrack_helper_put(ct->helper);
1047
1048 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1049         if (cda[CTA_MARK-1])
1050                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1051 #endif
1052
1053         DEBUGP("conntrack with id %u inserted\n", ct->id);
1054         return 0;
1055
1056 err:
1057         ip_conntrack_free(ct);
1058         return err;
1059 }
1060
1061 static int
1062 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1063                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1064 {
1065         struct ip_conntrack_tuple otuple, rtuple;
1066         struct ip_conntrack_tuple_hash *h = NULL;
1067         int err = 0;
1068
1069         DEBUGP("entered %s\n", __FUNCTION__);
1070
1071         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1072                 return -EINVAL;
1073
1074         if (cda[CTA_TUPLE_ORIG-1]) {
1075                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1076                 if (err < 0)
1077                         return err;
1078         }
1079
1080         if (cda[CTA_TUPLE_REPLY-1]) {
1081                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1082                 if (err < 0)
1083                         return err;
1084         }
1085
1086         write_lock_bh(&ip_conntrack_lock);
1087         if (cda[CTA_TUPLE_ORIG-1])
1088                 h = __ip_conntrack_find(&otuple, NULL);
1089         else if (cda[CTA_TUPLE_REPLY-1])
1090                 h = __ip_conntrack_find(&rtuple, NULL);
1091
1092         if (h == NULL) {
1093                 write_unlock_bh(&ip_conntrack_lock);
1094                 DEBUGP("no such conntrack, create new\n");
1095                 err = -ENOENT;
1096                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1097                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1098                 return err;
1099         }
1100         /* implicit 'else' */
1101
1102         /* we only allow nat config for new conntracks */
1103         if (cda[CTA_NAT-1]) {
1104                 err = -EINVAL;
1105                 goto out_unlock;
1106         }
1107
1108         /* We manipulate the conntrack inside the global conntrack table lock,
1109          * so there's no need to increase the refcount */
1110         DEBUGP("conntrack found\n");
1111         err = -EEXIST;
1112         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1113                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1114
1115 out_unlock:
1116         write_unlock_bh(&ip_conntrack_lock);
1117         return err;
1118 }
1119
1120 /***********************************************************************
1121  * EXPECT
1122  ***********************************************************************/
1123
1124 static inline int
1125 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1126                          const struct ip_conntrack_tuple *tuple,
1127                          enum ctattr_expect type)
1128 {
1129         struct nfattr *nest_parms = NFA_NEST(skb, type);
1130
1131         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1132                 goto nfattr_failure;
1133
1134         NFA_NEST_END(skb, nest_parms);
1135
1136         return 0;
1137
1138 nfattr_failure:
1139         return -1;
1140 }
1141
1142 static inline int
1143 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1144                           const struct ip_conntrack_expect *exp)
1145 {
1146         struct ip_conntrack *master = exp->master;
1147         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1148         u_int32_t id = htonl(exp->id);
1149
1150         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1151                 goto nfattr_failure;
1152         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1153                 goto nfattr_failure;
1154         if (ctnetlink_exp_dump_tuple(skb,
1155                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1156                                  CTA_EXPECT_MASTER) < 0)
1157                 goto nfattr_failure;
1158
1159         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1160         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1161
1162         return 0;
1163
1164 nfattr_failure:
1165         return -1;
1166 }
1167
1168 static int
1169 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1170                     int event,
1171                     int nowait,
1172                     const struct ip_conntrack_expect *exp)
1173 {
1174         struct nlmsghdr *nlh;
1175         struct nfgenmsg *nfmsg;
1176         unsigned char *b;
1177
1178         b = skb->tail;
1179
1180         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1181         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1182         nfmsg  = NLMSG_DATA(nlh);
1183
1184         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1185         nfmsg->nfgen_family = AF_INET;
1186         nfmsg->version      = NFNETLINK_V0;
1187         nfmsg->res_id       = 0;
1188
1189         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1190                 goto nfattr_failure;
1191
1192         nlh->nlmsg_len = skb->tail - b;
1193         return skb->len;
1194
1195 nlmsg_failure:
1196 nfattr_failure:
1197         skb_trim(skb, b - skb->data);
1198         return -1;
1199 }
1200
1201 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1202 static int ctnetlink_expect_event(struct notifier_block *this,
1203                                   unsigned long events, void *ptr)
1204 {
1205         struct nlmsghdr *nlh;
1206         struct nfgenmsg *nfmsg;
1207         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1208         struct sk_buff *skb;
1209         unsigned int type;
1210         unsigned char *b;
1211         int flags = 0;
1212         u16 proto;
1213
1214         if (events & IPEXP_NEW) {
1215                 type = IPCTNL_MSG_EXP_NEW;
1216                 flags = NLM_F_CREATE|NLM_F_EXCL;
1217         } else
1218                 return NOTIFY_DONE;
1219
1220         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1221         if (!skb)
1222                 return NOTIFY_DONE;
1223
1224         b = skb->tail;
1225
1226         type |= NFNL_SUBSYS_CTNETLINK << 8;
1227         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1228         nfmsg = NLMSG_DATA(nlh);
1229
1230         nlh->nlmsg_flags    = flags;
1231         nfmsg->nfgen_family = AF_INET;
1232         nfmsg->version      = NFNETLINK_V0;
1233         nfmsg->res_id       = 0;
1234
1235         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1236                 goto nfattr_failure;
1237
1238         nlh->nlmsg_len = skb->tail - b;
1239         proto = exp->tuple.dst.protonum;
1240         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1241         return NOTIFY_DONE;
1242
1243 nlmsg_failure:
1244 nfattr_failure:
1245         kfree_skb(skb);
1246         return NOTIFY_DONE;
1247 }
1248 #endif
1249
1250 static int
1251 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1252 {
1253         struct ip_conntrack_expect *exp = NULL;
1254         struct list_head *i;
1255         u_int32_t *id = (u_int32_t *) &cb->args[0];
1256
1257         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1258
1259         read_lock_bh(&ip_conntrack_lock);
1260         list_for_each_prev(i, &ip_conntrack_expect_list) {
1261                 exp = (struct ip_conntrack_expect *) i;
1262                 if (exp->id <= *id)
1263                         continue;
1264                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1265                                             cb->nlh->nlmsg_seq,
1266                                             IPCTNL_MSG_EXP_NEW,
1267                                             1, exp) < 0)
1268                         goto out;
1269                 *id = exp->id;
1270         }
1271 out:
1272         read_unlock_bh(&ip_conntrack_lock);
1273
1274         DEBUGP("leaving, last id=%llu\n", *id);
1275
1276         return skb->len;
1277 }
1278
1279 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1280         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1281         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1282 };
1283
1284 static int
1285 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1286                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1287 {
1288         struct ip_conntrack_tuple tuple;
1289         struct ip_conntrack_expect *exp;
1290         struct sk_buff *skb2;
1291         int err = 0;
1292
1293         DEBUGP("entered %s\n", __FUNCTION__);
1294
1295         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1296                 return -EINVAL;
1297
1298         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1299                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1300                 u32 rlen;
1301
1302                 if (msg->nfgen_family != AF_INET)
1303                         return -EAFNOSUPPORT;
1304
1305                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1306                                                 ctnetlink_exp_dump_table,
1307                                                 ctnetlink_done)) != 0)
1308                         return -EINVAL;
1309                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1310                 if (rlen > skb->len)
1311                         rlen = skb->len;
1312                 skb_pull(skb, rlen);
1313                 return 0;
1314         }
1315
1316         if (cda[CTA_EXPECT_MASTER-1])
1317                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1318         else
1319                 return -EINVAL;
1320
1321         if (err < 0)
1322                 return err;
1323
1324         exp = ip_conntrack_expect_find(&tuple);
1325         if (!exp)
1326                 return -ENOENT;
1327
1328         if (cda[CTA_EXPECT_ID-1]) {
1329                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1330                 if (exp->id != ntohl(id)) {
1331                         ip_conntrack_expect_put(exp);
1332                         return -ENOENT;
1333                 }
1334         }
1335
1336         err = -ENOMEM;
1337         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1338         if (!skb2)
1339                 goto out;
1340         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1341
1342         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1343                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1344                                       1, exp);
1345         if (err <= 0)
1346                 goto free;
1347
1348         ip_conntrack_expect_put(exp);
1349
1350         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1351
1352 free:
1353         kfree_skb(skb2);
1354 out:
1355         ip_conntrack_expect_put(exp);
1356         return err;
1357 }
1358
1359 static int
1360 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1361                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1362 {
1363         struct ip_conntrack_expect *exp, *tmp;
1364         struct ip_conntrack_tuple tuple;
1365         struct ip_conntrack_helper *h;
1366         int err;
1367
1368         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1369                 return -EINVAL;
1370
1371         if (cda[CTA_EXPECT_TUPLE-1]) {
1372                 /* delete a single expect by tuple */
1373                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1374                 if (err < 0)
1375                         return err;
1376
1377                 /* bump usage count to 2 */
1378                 exp = ip_conntrack_expect_find(&tuple);
1379                 if (!exp)
1380                         return -ENOENT;
1381
1382                 if (cda[CTA_EXPECT_ID-1]) {
1383                         u_int32_t id =
1384                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1385                         if (exp->id != ntohl(id)) {
1386                                 ip_conntrack_expect_put(exp);
1387                                 return -ENOENT;
1388                         }
1389                 }
1390
1391                 /* after list removal, usage count == 1 */
1392                 ip_conntrack_unexpect_related(exp);
1393                 /* have to put what we 'get' above.
1394                  * after this line usage count == 0 */
1395                 ip_conntrack_expect_put(exp);
1396         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1397                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1398
1399                 /* delete all expectations for this helper */
1400                 write_lock_bh(&ip_conntrack_lock);
1401                 h = __ip_conntrack_helper_find_byname(name);
1402                 if (!h) {
1403                         write_unlock_bh(&ip_conntrack_lock);
1404                         return -EINVAL;
1405                 }
1406                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1407                                          list) {
1408                         if (exp->master->helper == h
1409                             && del_timer(&exp->timeout)) {
1410                                 ip_ct_unlink_expect(exp);
1411                                 ip_conntrack_expect_put(exp);
1412                         }
1413                 }
1414                 write_unlock_bh(&ip_conntrack_lock);
1415         } else {
1416                 /* This basically means we have to flush everything*/
1417                 write_lock_bh(&ip_conntrack_lock);
1418                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1419                                          list) {
1420                         if (del_timer(&exp->timeout)) {
1421                                 ip_ct_unlink_expect(exp);
1422                                 ip_conntrack_expect_put(exp);
1423                         }
1424                 }
1425                 write_unlock_bh(&ip_conntrack_lock);
1426         }
1427
1428         return 0;
1429 }
1430 static int
1431 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1432 {
1433         return -EOPNOTSUPP;
1434 }
1435
1436 static int
1437 ctnetlink_create_expect(struct nfattr *cda[])
1438 {
1439         struct ip_conntrack_tuple tuple, mask, master_tuple;
1440         struct ip_conntrack_tuple_hash *h = NULL;
1441         struct ip_conntrack_expect *exp;
1442         struct ip_conntrack *ct;
1443         int err = 0;
1444
1445         DEBUGP("entered %s\n", __FUNCTION__);
1446
1447         /* caller guarantees that those three CTA_EXPECT_* exist */
1448         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1449         if (err < 0)
1450                 return err;
1451         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1452         if (err < 0)
1453                 return err;
1454         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1455         if (err < 0)
1456                 return err;
1457
1458         /* Look for master conntrack of this expectation */
1459         h = ip_conntrack_find_get(&master_tuple, NULL);
1460         if (!h)
1461                 return -ENOENT;
1462         ct = tuplehash_to_ctrack(h);
1463
1464         if (!ct->helper) {
1465                 /* such conntrack hasn't got any helper, abort */
1466                 err = -EINVAL;
1467                 goto out;
1468         }
1469
1470         exp = ip_conntrack_expect_alloc(ct);
1471         if (!exp) {
1472                 err = -ENOMEM;
1473                 goto out;
1474         }
1475
1476         exp->expectfn = NULL;
1477         exp->flags = 0;
1478         exp->master = ct;
1479         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1480         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1481
1482         err = ip_conntrack_expect_related(exp);
1483         ip_conntrack_expect_put(exp);
1484
1485 out:
1486         ip_conntrack_put(tuplehash_to_ctrack(h));
1487         return err;
1488 }
1489
1490 static int
1491 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1492                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1493 {
1494         struct ip_conntrack_tuple tuple;
1495         struct ip_conntrack_expect *exp;
1496         int err = 0;
1497
1498         DEBUGP("entered %s\n", __FUNCTION__);
1499
1500         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1501                 return -EINVAL;
1502
1503         if (!cda[CTA_EXPECT_TUPLE-1]
1504             || !cda[CTA_EXPECT_MASK-1]
1505             || !cda[CTA_EXPECT_MASTER-1])
1506                 return -EINVAL;
1507
1508         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1509         if (err < 0)
1510                 return err;
1511
1512         write_lock_bh(&ip_conntrack_lock);
1513         exp = __ip_conntrack_expect_find(&tuple);
1514
1515         if (!exp) {
1516                 write_unlock_bh(&ip_conntrack_lock);
1517                 err = -ENOENT;
1518                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1519                         err = ctnetlink_create_expect(cda);
1520                 return err;
1521         }
1522
1523         err = -EEXIST;
1524         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1525                 err = ctnetlink_change_expect(exp, cda);
1526         write_unlock_bh(&ip_conntrack_lock);
1527
1528         DEBUGP("leaving\n");
1529
1530         return err;
1531 }
1532
1533 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1534 static struct notifier_block ctnl_notifier = {
1535         .notifier_call  = ctnetlink_conntrack_event,
1536 };
1537
1538 static struct notifier_block ctnl_notifier_exp = {
1539         .notifier_call  = ctnetlink_expect_event,
1540 };
1541 #endif
1542
1543 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1544         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1545                                             .attr_count = CTA_MAX, },
1546         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1547                                             .attr_count = CTA_MAX, },
1548         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1549                                             .attr_count = CTA_MAX, },
1550         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1551                                             .attr_count = CTA_MAX, },
1552 };
1553
1554 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1555         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1556                                             .attr_count = CTA_EXPECT_MAX, },
1557         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1558                                             .attr_count = CTA_EXPECT_MAX, },
1559         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1560                                             .attr_count = CTA_EXPECT_MAX, },
1561 };
1562
1563 static struct nfnetlink_subsystem ctnl_subsys = {
1564         .name                           = "conntrack",
1565         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1566         .cb_count                       = IPCTNL_MSG_MAX,
1567         .cb                             = ctnl_cb,
1568 };
1569
1570 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1571         .name                           = "conntrack_expect",
1572         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1573         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1574         .cb                             = ctnl_exp_cb,
1575 };
1576
1577 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1578
1579 static int __init ctnetlink_init(void)
1580 {
1581         int ret;
1582
1583         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1584         ret = nfnetlink_subsys_register(&ctnl_subsys);
1585         if (ret < 0) {
1586                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1587                 goto err_out;
1588         }
1589
1590         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1591         if (ret < 0) {
1592                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1593                 goto err_unreg_subsys;
1594         }
1595
1596 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1597         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1598         if (ret < 0) {
1599                 printk("ctnetlink_init: cannot register notifier.\n");
1600                 goto err_unreg_exp_subsys;
1601         }
1602
1603         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1604         if (ret < 0) {
1605                 printk("ctnetlink_init: cannot expect register notifier.\n");
1606                 goto err_unreg_notifier;
1607         }
1608 #endif
1609
1610         return 0;
1611
1612 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1613 err_unreg_notifier:
1614         ip_conntrack_unregister_notifier(&ctnl_notifier);
1615 err_unreg_exp_subsys:
1616         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1617 #endif
1618 err_unreg_subsys:
1619         nfnetlink_subsys_unregister(&ctnl_subsys);
1620 err_out:
1621         return ret;
1622 }
1623
1624 static void __exit ctnetlink_exit(void)
1625 {
1626         printk("ctnetlink: unregistering from nfnetlink.\n");
1627
1628 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1629         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1630         ip_conntrack_unregister_notifier(&ctnl_notifier);
1631 #endif
1632
1633         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1634         nfnetlink_subsys_unregister(&ctnl_subsys);
1635         return;
1636 }
1637
1638 module_init(ctnetlink_init);
1639 module_exit(ctnetlink_exit);