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_int16_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_int16_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                 ip_conntrack_put(ct);
 733                 ct->timeout.function((unsigned long)ct);
 734                 return 0;
 735         }
 736         ip_conntrack_put(ct);
 737         DEBUGP("leaving\n");
 738
 739         return 0;
 740 }
 741
 742 static int
 743 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 744                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 745 {
 746         struct ip_conntrack_tuple_hash *h;
 747         struct ip_conntrack_tuple tuple;
 748         struct ip_conntrack *ct;
 749         struct sk_buff *skb2 = NULL;
 750         int err = 0;
 751
 752         DEBUGP("entered %s\n", __FUNCTION__);
 753
 754         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 755                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 756                 u32 rlen;
 757
 758                 if (msg->nfgen_family != AF_INET)
 759                         return -EAFNOSUPPORT;
 760
 761                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 762                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 763 #ifdef CONFIG_IP_NF_CT_ACCT
 764                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 765                                                 ctnetlink_dump_table_w,
 766                                                 ctnetlink_done)) != 0)
 767                                 return -EINVAL;
 768 #else
 769                         return -ENOTSUPP;
 770 #endif
 771                 } else {
 772                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 773                                                         ctnetlink_dump_table,
 774                                                         ctnetlink_done)) != 0)
 775                         return -EINVAL;
 776                 }
 777
 778                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 779                 if (rlen > skb->len)
 780                         rlen = skb->len;
 781                 skb_pull(skb, rlen);
 782                 return 0;
 783         }
 784
 785         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 786                 return -EINVAL;
 787
 788         if (cda[CTA_TUPLE_ORIG-1])
 789                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 790         else if (cda[CTA_TUPLE_REPLY-1])
 791                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 792         else
 793                 return -EINVAL;
 794
 795         if (err < 0)
 796                 return err;
 797
 798         h = ip_conntrack_find_get(&tuple, NULL);
 799         if (!h) {
 800                 DEBUGP("tuple not found in conntrack hash");
 801                 return -ENOENT;
 802         }
 803         DEBUGP("tuple found\n");
 804         ct = tuplehash_to_ctrack(h);
 805
 806         err = -ENOMEM;
 807         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 808         if (!skb2) {
 809                 ip_conntrack_put(ct);
 810                 return -ENOMEM;
 811         }
 812         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 813
 814         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 815                                   IPCTNL_MSG_CT_NEW, 1, ct);
 816         ip_conntrack_put(ct);
 817         if (err <= 0)
 818                 goto free;
 819
 820         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 821         if (err < 0)
 822                 goto out;
 823
 824         DEBUGP("leaving\n");
 825         return 0;
 826
 827 free:
 828         kfree_skb(skb2);
 829 out:
 830         return err;
 831 }
 832
 833 static inline int
 834 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 835 {
 836         unsigned long d;
 837         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 838         d = ct->status ^ status;
 839
 840         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 841                 /* unchangeable */
 842                 return -EINVAL;
 843
 844         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 845                 /* SEEN_REPLY bit can only be set */
 846                 return -EINVAL;
 847
 848
 849         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 850                 /* ASSURED bit can only be set */
 851                 return -EINVAL;
 852
 853         if (cda[CTA_NAT-1]) {
 854 #ifndef CONFIG_IP_NF_NAT_NEEDED
 855                 return -EINVAL;
 856 #else
 857                 unsigned int hooknum;
 858                 struct ip_nat_range range;
 859
 860                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 861                         return -EINVAL;
 862
 863                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 864                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 865                        htons(range.min.all), htons(range.max.all));
 866
 867                 /* This is tricky but it works. ip_nat_setup_info needs the
 868                  * hook number as parameter, so let's do the correct
 869                  * conversion and run away */
 870                 if (status & IPS_SRC_NAT_DONE)
 871                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 872                 else if (status & IPS_DST_NAT_DONE)
 873                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 874                 else
 875                         return -EINVAL; /* Missing NAT flags */
 876
 877                 DEBUGP("NAT status: %lu\n",
 878                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 879
 880                 if (ip_nat_initialized(ct, hooknum))
 881                         return -EEXIST;
 882                 ip_nat_setup_info(ct, &range, hooknum);
 883
 884                 DEBUGP("NAT status after setup_info: %lu\n",
 885                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 886 #endif
 887         }
 888
 889         /* Be careful here, modifying NAT bits can screw up things,
 890          * so don't let users modify them directly if they don't pass
 891          * ip_nat_range. */
 892         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 893         return 0;
 894 }
 895
 896
 897 static inline int
 898 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 899 {
 900         struct ip_conntrack_helper *helper;
 901         char *helpname;
 902         int err;
 903
 904         DEBUGP("entered %s\n", __FUNCTION__);
 905
 906         /* don't change helper of sibling connections */
 907         if (ct->master)
 908                 return -EINVAL;
 909
 910         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 911         if (err < 0)
 912                 return err;
 913
 914         helper = __ip_conntrack_helper_find_byname(helpname);
 915         if (!helper) {
 916                 if (!strcmp(helpname, ""))
 917                         helper = NULL;
 918                 else
 919                         return -EINVAL;
 920         }
 921
 922         if (ct->helper) {
 923                 if (!helper) {
 924                         /* we had a helper before ... */
 925                         ip_ct_remove_expectations(ct);
 926                         ct->helper = NULL;
 927                 } else {
 928                         /* need to zero data of old helper */
 929                         memset(&ct->help, 0, sizeof(ct->help));
 930                 }
 931         }
 932
 933         ct->helper = helper;
 934
 935         return 0;
 936 }
 937
 938 static inline int
 939 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 940 {
 941         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 942
 943         if (!del_timer(&ct->timeout))
 944                 return -ETIME;
 945
 946         ct->timeout.expires = jiffies + timeout * HZ;
 947         add_timer(&ct->timeout);
 948
 949         return 0;
 950 }
 951
 952 static inline int
 953 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 954 {
 955         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 956         struct ip_conntrack_protocol *proto;
 957         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 958         int err = 0;
 959
 960         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 961
 962         proto = ip_conntrack_proto_find_get(npt);
 963
 964         if (proto->from_nfattr)
 965                 err = proto->from_nfattr(tb, ct);
 966         ip_conntrack_proto_put(proto);
 967
 968         return err;
 969 }
 970
 971 static int
 972 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 973 {
 974         int err;
 975
 976         DEBUGP("entered %s\n", __FUNCTION__);
 977
 978         if (cda[CTA_HELP-1]) {
 979                 err = ctnetlink_change_helper(ct, cda);
 980                 if (err < 0)
 981                         return err;
 982         }
 983
 984         if (cda[CTA_TIMEOUT-1]) {
 985                 err = ctnetlink_change_timeout(ct, cda);
 986                 if (err < 0)
 987                         return err;
 988         }
 989
 990         if (cda[CTA_STATUS-1]) {
 991                 err = ctnetlink_change_status(ct, cda);
 992                 if (err < 0)
 993                         return err;
 994         }
 995
 996         if (cda[CTA_PROTOINFO-1]) {
 997                 err = ctnetlink_change_protoinfo(ct, cda);
 998                 if (err < 0)
 999                         return err;
1000         }
1001
1002 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1003         if (cda[CTA_MARK-1])
1004                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1005 #endif
1006
1007         DEBUGP("all done\n");
1008         return 0;
1009 }
1010
1011 static int
1012 ctnetlink_create_conntrack(struct nfattr *cda[],
1013                            struct ip_conntrack_tuple *otuple,
1014                            struct ip_conntrack_tuple *rtuple)
1015 {
1016         struct ip_conntrack *ct;
1017         int err = -EINVAL;
1018
1019         DEBUGP("entered %s\n", __FUNCTION__);
1020
1021         ct = ip_conntrack_alloc(otuple, rtuple);
1022         if (ct == NULL || IS_ERR(ct))
1023                 return -ENOMEM;
1024
1025         if (!cda[CTA_TIMEOUT-1])
1026                 goto err;
1027         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1028
1029         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1030         ct->status |= IPS_CONFIRMED;
1031
1032         err = ctnetlink_change_status(ct, cda);
1033         if (err < 0)
1034                 goto err;
1035
1036         if (cda[CTA_PROTOINFO-1]) {
1037                 err = ctnetlink_change_protoinfo(ct, cda);
1038                 if (err < 0)
1039                         return err;
1040         }
1041
1042         ct->helper = ip_conntrack_helper_find_get(rtuple);
1043
1044         add_timer(&ct->timeout);
1045         ip_conntrack_hash_insert(ct);
1046
1047         if (ct->helper)
1048                 ip_conntrack_helper_put(ct->helper);
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         DEBUGP("conntrack with id %u inserted\n", ct->id);
1056         return 0;
1057
1058 err:
1059         ip_conntrack_free(ct);
1060         return err;
1061 }
1062
1063 static int
1064 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1065                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1066 {
1067         struct ip_conntrack_tuple otuple, rtuple;
1068         struct ip_conntrack_tuple_hash *h = NULL;
1069         int err = 0;
1070
1071         DEBUGP("entered %s\n", __FUNCTION__);
1072
1073         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1074                 return -EINVAL;
1075
1076         if (cda[CTA_TUPLE_ORIG-1]) {
1077                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1078                 if (err < 0)
1079                         return err;
1080         }
1081
1082         if (cda[CTA_TUPLE_REPLY-1]) {
1083                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1084                 if (err < 0)
1085                         return err;
1086         }
1087
1088         write_lock_bh(&ip_conntrack_lock);
1089         if (cda[CTA_TUPLE_ORIG-1])
1090                 h = __ip_conntrack_find(&otuple, NULL);
1091         else if (cda[CTA_TUPLE_REPLY-1])
1092                 h = __ip_conntrack_find(&rtuple, NULL);
1093
1094         if (h == NULL) {
1095                 write_unlock_bh(&ip_conntrack_lock);
1096                 DEBUGP("no such conntrack, create new\n");
1097                 err = -ENOENT;
1098                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1099                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1100                 return err;
1101         }
1102         /* implicit 'else' */
1103
1104         /* we only allow nat config for new conntracks */
1105         if (cda[CTA_NAT-1]) {
1106                 err = -EINVAL;
1107                 goto out_unlock;
1108         }
1109
1110         /* We manipulate the conntrack inside the global conntrack table lock,
1111          * so there's no need to increase the refcount */
1112         DEBUGP("conntrack found\n");
1113         err = -EEXIST;
1114         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1115                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1116
1117 out_unlock:
1118         write_unlock_bh(&ip_conntrack_lock);
1119         return err;
1120 }
1121
1122 /***********************************************************************
1123  * EXPECT
1124  ***********************************************************************/
1125
1126 static inline int
1127 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1128                          const struct ip_conntrack_tuple *tuple,
1129                          enum ctattr_expect type)
1130 {
1131         struct nfattr *nest_parms = NFA_NEST(skb, type);
1132
1133         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1134                 goto nfattr_failure;
1135
1136         NFA_NEST_END(skb, nest_parms);
1137
1138         return 0;
1139
1140 nfattr_failure:
1141         return -1;
1142 }
1143
1144 static inline int
1145 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1146                           const struct ip_conntrack_expect *exp)
1147 {
1148         struct ip_conntrack *master = exp->master;
1149         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1150         u_int32_t id = htonl(exp->id);
1151
1152         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1153                 goto nfattr_failure;
1154         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1155                 goto nfattr_failure;
1156         if (ctnetlink_exp_dump_tuple(skb,
1157                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1158                                  CTA_EXPECT_MASTER) < 0)
1159                 goto nfattr_failure;
1160
1161         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1162         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1163
1164         return 0;
1165
1166 nfattr_failure:
1167         return -1;
1168 }
1169
1170 static int
1171 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1172                     int event,
1173                     int nowait,
1174                     const struct ip_conntrack_expect *exp)
1175 {
1176         struct nlmsghdr *nlh;
1177         struct nfgenmsg *nfmsg;
1178         unsigned char *b;
1179
1180         b = skb->tail;
1181
1182         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1183         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1184         nfmsg  = NLMSG_DATA(nlh);
1185
1186         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1187         nfmsg->nfgen_family = AF_INET;
1188         nfmsg->version      = NFNETLINK_V0;
1189         nfmsg->res_id       = 0;
1190
1191         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1192                 goto nfattr_failure;
1193
1194         nlh->nlmsg_len = skb->tail - b;
1195         return skb->len;
1196
1197 nlmsg_failure:
1198 nfattr_failure:
1199         skb_trim(skb, b - skb->data);
1200         return -1;
1201 }
1202
1203 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1204 static int ctnetlink_expect_event(struct notifier_block *this,
1205                                   unsigned long events, void *ptr)
1206 {
1207         struct nlmsghdr *nlh;
1208         struct nfgenmsg *nfmsg;
1209         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1210         struct sk_buff *skb;
1211         unsigned int type;
1212         unsigned char *b;
1213         int flags = 0;
1214         u16 proto;
1215
1216         if (events & IPEXP_NEW) {
1217                 type = IPCTNL_MSG_EXP_NEW;
1218                 flags = NLM_F_CREATE|NLM_F_EXCL;
1219         } else
1220                 return NOTIFY_DONE;
1221
1222         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1223         if (!skb)
1224                 return NOTIFY_DONE;
1225
1226         b = skb->tail;
1227
1228         type |= NFNL_SUBSYS_CTNETLINK << 8;
1229         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1230         nfmsg = NLMSG_DATA(nlh);
1231
1232         nlh->nlmsg_flags    = flags;
1233         nfmsg->nfgen_family = AF_INET;
1234         nfmsg->version      = NFNETLINK_V0;
1235         nfmsg->res_id       = 0;
1236
1237         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1238                 goto nfattr_failure;
1239
1240         nlh->nlmsg_len = skb->tail - b;
1241         proto = exp->tuple.dst.protonum;
1242         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1243         return NOTIFY_DONE;
1244
1245 nlmsg_failure:
1246 nfattr_failure:
1247         kfree_skb(skb);
1248         return NOTIFY_DONE;
1249 }
1250 #endif
1251
1252 static int
1253 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1254 {
1255         struct ip_conntrack_expect *exp = NULL;
1256         struct list_head *i;
1257         u_int32_t *id = (u_int32_t *) &cb->args[0];
1258
1259         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1260
1261         read_lock_bh(&ip_conntrack_lock);
1262         list_for_each_prev(i, &ip_conntrack_expect_list) {
1263                 exp = (struct ip_conntrack_expect *) i;
1264                 if (exp->id <= *id)
1265                         continue;
1266                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1267                                             cb->nlh->nlmsg_seq,
1268                                             IPCTNL_MSG_EXP_NEW,
1269                                             1, exp) < 0)
1270                         goto out;
1271                 *id = exp->id;
1272         }
1273 out:
1274         read_unlock_bh(&ip_conntrack_lock);
1275
1276         DEBUGP("leaving, last id=%llu\n", *id);
1277
1278         return skb->len;
1279 }
1280
1281 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1282         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1283         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1284 };
1285
1286 static int
1287 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1288                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1289 {
1290         struct ip_conntrack_tuple tuple;
1291         struct ip_conntrack_expect *exp;
1292         struct sk_buff *skb2;
1293         int err = 0;
1294
1295         DEBUGP("entered %s\n", __FUNCTION__);
1296
1297         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1298                 return -EINVAL;
1299
1300         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1301                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1302                 u32 rlen;
1303
1304                 if (msg->nfgen_family != AF_INET)
1305                         return -EAFNOSUPPORT;
1306
1307                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1308                                                 ctnetlink_exp_dump_table,
1309                                                 ctnetlink_done)) != 0)
1310                         return -EINVAL;
1311                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1312                 if (rlen > skb->len)
1313                         rlen = skb->len;
1314                 skb_pull(skb, rlen);
1315                 return 0;
1316         }
1317
1318         if (cda[CTA_EXPECT_MASTER-1])
1319                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1320         else
1321                 return -EINVAL;
1322
1323         if (err < 0)
1324                 return err;
1325
1326         exp = ip_conntrack_expect_find(&tuple);
1327         if (!exp)
1328                 return -ENOENT;
1329
1330         if (cda[CTA_EXPECT_ID-1]) {
1331                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1332                 if (exp->id != ntohl(id)) {
1333                         ip_conntrack_expect_put(exp);
1334                         return -ENOENT;
1335                 }
1336         }
1337
1338         err = -ENOMEM;
1339         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1340         if (!skb2)
1341                 goto out;
1342         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1343
1344         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1345                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1346                                       1, exp);
1347         if (err <= 0)
1348                 goto free;
1349
1350         ip_conntrack_expect_put(exp);
1351
1352         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1353
1354 free:
1355         kfree_skb(skb2);
1356 out:
1357         ip_conntrack_expect_put(exp);
1358         return err;
1359 }
1360
1361 static int
1362 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1363                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1364 {
1365         struct ip_conntrack_expect *exp, *tmp;
1366         struct ip_conntrack_tuple tuple;
1367         struct ip_conntrack_helper *h;
1368         int err;
1369
1370         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1371                 return -EINVAL;
1372
1373         if (cda[CTA_EXPECT_TUPLE-1]) {
1374                 /* delete a single expect by tuple */
1375                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1376                 if (err < 0)
1377                         return err;
1378
1379                 /* bump usage count to 2 */
1380                 exp = ip_conntrack_expect_find(&tuple);
1381                 if (!exp)
1382                         return -ENOENT;
1383
1384                 if (cda[CTA_EXPECT_ID-1]) {
1385                         u_int32_t id =
1386                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1387                         if (exp->id != ntohl(id)) {
1388                                 ip_conntrack_expect_put(exp);
1389                                 return -ENOENT;
1390                         }
1391                 }
1392
1393                 /* after list removal, usage count == 1 */
1394                 ip_conntrack_unexpect_related(exp);
1395                 /* have to put what we 'get' above.
1396                  * after this line usage count == 0 */
1397                 ip_conntrack_expect_put(exp);
1398         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1399                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1400
1401                 /* delete all expectations for this helper */
1402                 write_lock_bh(&ip_conntrack_lock);
1403                 h = __ip_conntrack_helper_find_byname(name);
1404                 if (!h) {
1405                         write_unlock_bh(&ip_conntrack_lock);
1406                         return -EINVAL;
1407                 }
1408                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1409                                          list) {
1410                         if (exp->master->helper == h
1411                             && del_timer(&exp->timeout)) {
1412                                 ip_ct_unlink_expect(exp);
1413                                 ip_conntrack_expect_put(exp);
1414                         }
1415                 }
1416                 write_unlock_bh(&ip_conntrack_lock);
1417         } else {
1418                 /* This basically means we have to flush everything*/
1419                 write_lock_bh(&ip_conntrack_lock);
1420                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1421                                          list) {
1422                         if (del_timer(&exp->timeout)) {
1423                                 ip_ct_unlink_expect(exp);
1424                                 ip_conntrack_expect_put(exp);
1425                         }
1426                 }
1427                 write_unlock_bh(&ip_conntrack_lock);
1428         }
1429
1430         return 0;
1431 }
1432 static int
1433 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1434 {
1435         return -EOPNOTSUPP;
1436 }
1437
1438 static int
1439 ctnetlink_create_expect(struct nfattr *cda[])
1440 {
1441         struct ip_conntrack_tuple tuple, mask, master_tuple;
1442         struct ip_conntrack_tuple_hash *h = NULL;
1443         struct ip_conntrack_expect *exp;
1444         struct ip_conntrack *ct;
1445         int err = 0;
1446
1447         DEBUGP("entered %s\n", __FUNCTION__);
1448
1449         /* caller guarantees that those three CTA_EXPECT_* exist */
1450         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1451         if (err < 0)
1452                 return err;
1453         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1454         if (err < 0)
1455                 return err;
1456         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1457         if (err < 0)
1458                 return err;
1459
1460         /* Look for master conntrack of this expectation */
1461         h = ip_conntrack_find_get(&master_tuple, NULL);
1462         if (!h)
1463                 return -ENOENT;
1464         ct = tuplehash_to_ctrack(h);
1465
1466         if (!ct->helper) {
1467                 /* such conntrack hasn't got any helper, abort */
1468                 err = -EINVAL;
1469                 goto out;
1470         }
1471
1472         exp = ip_conntrack_expect_alloc(ct);
1473         if (!exp) {
1474                 err = -ENOMEM;
1475                 goto out;
1476         }
1477
1478         exp->expectfn = NULL;
1479         exp->flags = 0;
1480         exp->master = ct;
1481         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1482         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1483
1484         err = ip_conntrack_expect_related(exp);
1485         ip_conntrack_expect_put(exp);
1486
1487 out:
1488         ip_conntrack_put(tuplehash_to_ctrack(h));
1489         return err;
1490 }
1491
1492 static int
1493 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1494                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1495 {
1496         struct ip_conntrack_tuple tuple;
1497         struct ip_conntrack_expect *exp;
1498         int err = 0;
1499
1500         DEBUGP("entered %s\n", __FUNCTION__);
1501
1502         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1503                 return -EINVAL;
1504
1505         if (!cda[CTA_EXPECT_TUPLE-1]
1506             || !cda[CTA_EXPECT_MASK-1]
1507             || !cda[CTA_EXPECT_MASTER-1])
1508                 return -EINVAL;
1509
1510         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1511         if (err < 0)
1512                 return err;
1513
1514         write_lock_bh(&ip_conntrack_lock);
1515         exp = __ip_conntrack_expect_find(&tuple);
1516
1517         if (!exp) {
1518                 write_unlock_bh(&ip_conntrack_lock);
1519                 err = -ENOENT;
1520                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1521                         err = ctnetlink_create_expect(cda);
1522                 return err;
1523         }
1524
1525         err = -EEXIST;
1526         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1527                 err = ctnetlink_change_expect(exp, cda);
1528         write_unlock_bh(&ip_conntrack_lock);
1529
1530         DEBUGP("leaving\n");
1531
1532         return err;
1533 }
1534
1535 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1536 static struct notifier_block ctnl_notifier = {
1537         .notifier_call  = ctnetlink_conntrack_event,
1538 };
1539
1540 static struct notifier_block ctnl_notifier_exp = {
1541         .notifier_call  = ctnetlink_expect_event,
1542 };
1543 #endif
1544
1545 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1546         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1547                                             .attr_count = CTA_MAX, },
1548         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1549                                             .attr_count = CTA_MAX, },
1550         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1551                                             .attr_count = CTA_MAX, },
1552         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1553                                             .attr_count = CTA_MAX, },
1554 };
1555
1556 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1557         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1558                                             .attr_count = CTA_EXPECT_MAX, },
1559         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1560                                             .attr_count = CTA_EXPECT_MAX, },
1561         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1562                                             .attr_count = CTA_EXPECT_MAX, },
1563 };
1564
1565 static struct nfnetlink_subsystem ctnl_subsys = {
1566         .name                           = "conntrack",
1567         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1568         .cb_count                       = IPCTNL_MSG_MAX,
1569         .cb                             = ctnl_cb,
1570 };
1571
1572 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1573         .name                           = "conntrack_expect",
1574         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1575         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1576         .cb                             = ctnl_exp_cb,
1577 };
1578
1579 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1580
1581 static int __init ctnetlink_init(void)
1582 {
1583         int ret;
1584
1585         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1586         ret = nfnetlink_subsys_register(&ctnl_subsys);
1587         if (ret < 0) {
1588                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1589                 goto err_out;
1590         }
1591
1592         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1593         if (ret < 0) {
1594                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1595                 goto err_unreg_subsys;
1596         }
1597
1598 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1599         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1600         if (ret < 0) {
1601                 printk("ctnetlink_init: cannot register notifier.\n");
1602                 goto err_unreg_exp_subsys;
1603         }
1604
1605         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1606         if (ret < 0) {
1607                 printk("ctnetlink_init: cannot expect register notifier.\n");
1608                 goto err_unreg_notifier;
1609         }
1610 #endif
1611
1612         return 0;
1613
1614 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1615 err_unreg_notifier:
1616         ip_conntrack_unregister_notifier(&ctnl_notifier);
1617 err_unreg_exp_subsys:
1618         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1619 #endif
1620 err_unreg_subsys:
1621         nfnetlink_subsys_unregister(&ctnl_subsys);
1622 err_out:
1623         return ret;
1624 }
1625
1626 static void __exit ctnetlink_exit(void)
1627 {
1628         printk("ctnetlink: unregistering from nfnetlink.\n");
1629
1630 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1631         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1632         ip_conntrack_unregister_notifier(&ctnl_notifier);
1633 #endif
1634
1635         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1636         nfnetlink_subsys_unregister(&ctnl_subsys);
1637         return;
1638 }
1639
1640 module_init(ctnetlink_init);
1641 module_exit(ctnetlink_exit);