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/notifier.h>
  31 #include <linux/rtnetlink.h>
  32
  33 #include <linux/netfilter.h>
  34 #include <linux/netfilter_ipv4.h>
  35 #include <linux/netfilter_ipv4/ip_tables.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  38 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  39 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  40 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  41
  42 #include <linux/netfilter/nfnetlink.h>
  43 #include <linux/netfilter/nfnetlink_conntrack.h>
  44
  45 MODULE_LICENSE("GPL");
  46
  47 static char __initdata version[] = "0.90";
  48
  49 #if 0
  50 #define DEBUGP printk
  51 #else
  52 #define DEBUGP(format, args...)
  53 #endif
  54
  55
  56 static inline int
  57 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  58                             const struct ip_conntrack_tuple *tuple)
  59 {
  60         struct ip_conntrack_protocol *proto;
  61         int ret = 0;
  62
  63         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  64
  65         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  66         if (likely(proto && proto->tuple_to_nfattr)) {
  67                 ret = proto->tuple_to_nfattr(skb, tuple);
  68                 ip_conntrack_proto_put(proto);
  69         }
  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 || !proto->to_nfattr)
 136                 return 0;
 137
 138         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 139
 140         ret = proto->to_nfattr(skb, nest_proto, ct);
 141
 142         ip_conntrack_proto_put(proto);
 143
 144         NFA_NEST_END(skb, nest_proto);
 145
 146         return ret;
 147
 148 nfattr_failure:
 149         return -1;
 150 }
 151
 152 static inline int
 153 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 154 {
 155         struct nfattr *nest_helper;
 156
 157         if (!ct->helper)
 158                 return 0;
 159
 160         nest_helper = NFA_NEST(skb, CTA_HELP);
 161         NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
 162
 163         if (ct->helper->to_nfattr)
 164                 ct->helper->to_nfattr(skb, ct);
 165
 166         NFA_NEST_END(skb, nest_helper);
 167
 168         return 0;
 169
 170 nfattr_failure:
 171         return -1;
 172 }
 173
 174 #ifdef CONFIG_IP_NF_CT_ACCT
 175 static inline int
 176 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 177                         enum ip_conntrack_dir dir)
 178 {
 179         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 180         struct nfattr *nest_count = NFA_NEST(skb, type);
 181         u_int32_t tmp;
 182
 183         tmp = htonl(ct->counters[dir].packets);
 184         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 185
 186         tmp = htonl(ct->counters[dir].bytes);
 187         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 188
 189         NFA_NEST_END(skb, nest_count);
 190
 191         return 0;
 192
 193 nfattr_failure:
 194         return -1;
 195 }
 196 #else
 197 #define ctnetlink_dump_counters(a, b, c) (0)
 198 #endif
 199
 200 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 201 static inline int
 202 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 203 {
 204         u_int32_t mark = htonl(ct->mark);
 205
 206         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 207         return 0;
 208
 209 nfattr_failure:
 210         return -1;
 211 }
 212 #else
 213 #define ctnetlink_dump_mark(a, b) (0)
 214 #endif
 215
 216 static inline int
 217 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 218 {
 219         u_int32_t id = htonl(ct->id);
 220         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 221         return 0;
 222
 223 nfattr_failure:
 224         return -1;
 225 }
 226
 227 static inline int
 228 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 229 {
 230         unsigned int use = htonl(atomic_read(&ct->ct_general.use));
 231
 232         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 233         return 0;
 234
 235 nfattr_failure:
 236         return -1;
 237 }
 238
 239 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 240
 241 static int
 242 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 243                     int event, int nowait,
 244                     const struct ip_conntrack *ct)
 245 {
 246         struct nlmsghdr *nlh;
 247         struct nfgenmsg *nfmsg;
 248         struct nfattr *nest_parms;
 249         unsigned char *b;
 250
 251         b = skb->tail;
 252
 253         event |= NFNL_SUBSYS_CTNETLINK << 8;
 254         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 255         nfmsg  = NLMSG_DATA(nlh);
 256
 257         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 258         nfmsg->nfgen_family = AF_INET;
 259         nfmsg->version      = NFNETLINK_V0;
 260         nfmsg->res_id       = 0;
 261
 262         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 263         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 264                 goto nfattr_failure;
 265         NFA_NEST_END(skb, nest_parms);
 266
 267         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 268         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 269                 goto nfattr_failure;
 270         NFA_NEST_END(skb, nest_parms);
 271
 272         if (ctnetlink_dump_status(skb, ct) < 0 ||
 273             ctnetlink_dump_timeout(skb, ct) < 0 ||
 274             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 275             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 276             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 277             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 278             ctnetlink_dump_mark(skb, ct) < 0 ||
 279             ctnetlink_dump_id(skb, ct) < 0 ||
 280             ctnetlink_dump_use(skb, ct) < 0)
 281                 goto nfattr_failure;
 282
 283         nlh->nlmsg_len = skb->tail - b;
 284         return skb->len;
 285
 286 nlmsg_failure:
 287 nfattr_failure:
 288         skb_trim(skb, b - skb->data);
 289         return -1;
 290 }
 291
 292 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 293 static int ctnetlink_conntrack_event(struct notifier_block *this,
 294                                      unsigned long events, void *ptr)
 295 {
 296         struct nlmsghdr *nlh;
 297         struct nfgenmsg *nfmsg;
 298         struct nfattr *nest_parms;
 299         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 300         struct sk_buff *skb;
 301         unsigned int type;
 302         unsigned char *b;
 303         unsigned int flags = 0, group;
 304
 305         /* ignore our fake conntrack entry */
 306         if (ct == &ip_conntrack_untracked)
 307                 return NOTIFY_DONE;
 308
 309         if (events & IPCT_DESTROY) {
 310                 type = IPCTNL_MSG_CT_DELETE;
 311                 group = NFNLGRP_CONNTRACK_DESTROY;
 312                 goto alloc_skb;
 313         }
 314         if (events & (IPCT_NEW | IPCT_RELATED)) {
 315                 type = IPCTNL_MSG_CT_NEW;
 316                 flags = NLM_F_CREATE|NLM_F_EXCL;
 317                 /* dump everything */
 318                 events = ~0UL;
 319                 group = NFNLGRP_CONNTRACK_NEW;
 320                 goto alloc_skb;
 321         }
 322         if (events & (IPCT_STATUS |
 323                       IPCT_PROTOINFO |
 324                       IPCT_HELPER |
 325                       IPCT_HELPINFO |
 326                       IPCT_NATINFO)) {
 327                 type = IPCTNL_MSG_CT_NEW;
 328                 group = NFNLGRP_CONNTRACK_UPDATE;
 329                 goto alloc_skb;
 330         }
 331
 332         return NOTIFY_DONE;
 333
 334 alloc_skb:
 335   /* FIXME: Check if there are any listeners before, don't hurt performance */
 336
 337         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 338         if (!skb)
 339                 return NOTIFY_DONE;
 340
 341         b = skb->tail;
 342
 343         type |= NFNL_SUBSYS_CTNETLINK << 8;
 344         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 345         nfmsg = NLMSG_DATA(nlh);
 346
 347         nlh->nlmsg_flags    = flags;
 348         nfmsg->nfgen_family = AF_INET;
 349         nfmsg->version  = NFNETLINK_V0;
 350         nfmsg->res_id   = 0;
 351
 352         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 353         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 354                 goto nfattr_failure;
 355         NFA_NEST_END(skb, nest_parms);
 356
 357         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 358         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 359                 goto nfattr_failure;
 360         NFA_NEST_END(skb, nest_parms);
 361
 362         /* NAT stuff is now a status flag */
 363         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 364             && ctnetlink_dump_status(skb, ct) < 0)
 365                 goto nfattr_failure;
 366         if (events & IPCT_REFRESH
 367             && ctnetlink_dump_timeout(skb, ct) < 0)
 368                 goto nfattr_failure;
 369         if (events & IPCT_PROTOINFO
 370             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 371                 goto nfattr_failure;
 372         if (events & IPCT_HELPINFO
 373             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 374                 goto nfattr_failure;
 375
 376         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 377             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 378                 goto nfattr_failure;
 379
 380         nlh->nlmsg_len = skb->tail - b;
 381         nfnetlink_send(skb, 0, group, 0);
 382         return NOTIFY_DONE;
 383
 384 nlmsg_failure:
 385 nfattr_failure:
 386         kfree_skb(skb);
 387         return NOTIFY_DONE;
 388 }
 389 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 390
 391 static int ctnetlink_done(struct netlink_callback *cb)
 392 {
 393         DEBUGP("entered %s\n", __FUNCTION__);
 394         return 0;
 395 }
 396
 397 static int
 398 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 399 {
 400         struct ip_conntrack *ct = NULL;
 401         struct ip_conntrack_tuple_hash *h;
 402         struct list_head *i;
 403         u_int32_t *id = (u_int32_t *) &cb->args[1];
 404
 405         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 406                         cb->args[0], *id);
 407
 408         read_lock_bh(&ip_conntrack_lock);
 409         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 410                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 411                         h = (struct ip_conntrack_tuple_hash *) i;
 412                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 413                                 continue;
 414                         ct = tuplehash_to_ctrack(h);
 415                         if (ct->id <= *id)
 416                                 continue;
 417                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 418                                                 cb->nlh->nlmsg_seq,
 419                                                 IPCTNL_MSG_CT_NEW,
 420                                                 1, ct) < 0)
 421                                 goto out;
 422                         *id = ct->id;
 423                 }
 424         }
 425 out:
 426         read_unlock_bh(&ip_conntrack_lock);
 427
 428         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 429
 430         return skb->len;
 431 }
 432
 433 #ifdef CONFIG_IP_NF_CT_ACCT
 434 static int
 435 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 436 {
 437         struct ip_conntrack *ct = NULL;
 438         struct ip_conntrack_tuple_hash *h;
 439         struct list_head *i;
 440         u_int32_t *id = (u_int32_t *) &cb->args[1];
 441
 442         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 443                         cb->args[0], *id);
 444
 445         write_lock_bh(&ip_conntrack_lock);
 446         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 447                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 448                         h = (struct ip_conntrack_tuple_hash *) i;
 449                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 450                                 continue;
 451                         ct = tuplehash_to_ctrack(h);
 452                         if (ct->id <= *id)
 453                                 continue;
 454                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 455                                                 cb->nlh->nlmsg_seq,
 456                                                 IPCTNL_MSG_CT_NEW,
 457                                                 1, ct) < 0)
 458                                 goto out;
 459                         *id = ct->id;
 460
 461                         memset(&ct->counters, 0, sizeof(ct->counters));
 462                 }
 463         }
 464 out:
 465         write_unlock_bh(&ip_conntrack_lock);
 466
 467         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 468
 469         return skb->len;
 470 }
 471 #endif
 472
 473 static const int cta_min_ip[CTA_IP_MAX] = {
 474         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 475         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 476 };
 477
 478 static inline int
 479 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 480 {
 481         struct nfattr *tb[CTA_IP_MAX];
 482
 483         DEBUGP("entered %s\n", __FUNCTION__);
 484
 485         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 486
 487         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 488                 return -EINVAL;
 489
 490         if (!tb[CTA_IP_V4_SRC-1])
 491                 return -EINVAL;
 492         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 493
 494         if (!tb[CTA_IP_V4_DST-1])
 495                 return -EINVAL;
 496         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 497
 498         DEBUGP("leaving\n");
 499
 500         return 0;
 501 }
 502
 503 static const int cta_min_proto[CTA_PROTO_MAX] = {
 504         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 505         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 506         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 507         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 508         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 509         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 510 };
 511
 512 static inline int
 513 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 514                             struct ip_conntrack_tuple *tuple)
 515 {
 516         struct nfattr *tb[CTA_PROTO_MAX];
 517         struct ip_conntrack_protocol *proto;
 518         int ret = 0;
 519
 520         DEBUGP("entered %s\n", __FUNCTION__);
 521
 522         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 523
 524         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 525                 return -EINVAL;
 526
 527         if (!tb[CTA_PROTO_NUM-1])
 528                 return -EINVAL;
 529         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 530
 531         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 532
 533         if (likely(proto && proto->nfattr_to_tuple)) {
 534                 ret = proto->nfattr_to_tuple(tb, tuple);
 535                 ip_conntrack_proto_put(proto);
 536         }
 537
 538         return ret;
 539 }
 540
 541 static inline int
 542 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 543                       enum ctattr_tuple type)
 544 {
 545         struct nfattr *tb[CTA_TUPLE_MAX];
 546         int err;
 547
 548         DEBUGP("entered %s\n", __FUNCTION__);
 549
 550         memset(tuple, 0, sizeof(*tuple));
 551
 552         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 553
 554         if (!tb[CTA_TUPLE_IP-1])
 555                 return -EINVAL;
 556
 557         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 558         if (err < 0)
 559                 return err;
 560
 561         if (!tb[CTA_TUPLE_PROTO-1])
 562                 return -EINVAL;
 563
 564         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 565         if (err < 0)
 566                 return err;
 567
 568         /* orig and expect tuples get DIR_ORIGINAL */
 569         if (type == CTA_TUPLE_REPLY)
 570                 tuple->dst.dir = IP_CT_DIR_REPLY;
 571         else
 572                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 573
 574         DUMP_TUPLE(tuple);
 575
 576         DEBUGP("leaving\n");
 577
 578         return 0;
 579 }
 580
 581 #ifdef CONFIG_IP_NF_NAT_NEEDED
 582 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
 583         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 584         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 585 };
 586
 587 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 588                                      const struct ip_conntrack *ct,
 589                                      struct ip_nat_range *range)
 590 {
 591         struct nfattr *tb[CTA_PROTONAT_MAX];
 592         struct ip_nat_protocol *npt;
 593
 594         DEBUGP("entered %s\n", __FUNCTION__);
 595
 596         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 597
 598         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 599                 return -1;
 600
 601         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 602         if (!npt)
 603                 return 0;
 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 inline int
 621 ctnetlink_parse_nat(struct nfattr *cda[],
 622                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 623 {
 624         struct nfattr *tb[CTA_NAT_MAX];
 625         int err;
 626
 627         DEBUGP("entered %s\n", __FUNCTION__);
 628
 629         memset(range, 0, sizeof(*range));
 630
 631         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 632
 633         if (tb[CTA_NAT_MINIP-1])
 634                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 635
 636         if (!tb[CTA_NAT_MAXIP-1])
 637                 range->max_ip = range->min_ip;
 638         else
 639                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 640
 641         if (range->min_ip)
 642                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 643
 644         if (!tb[CTA_NAT_PROTO-1])
 645                 return 0;
 646
 647         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 648         if (err < 0)
 649                 return err;
 650
 651         DEBUGP("leaving\n");
 652         return 0;
 653 }
 654 #endif
 655
 656 static inline int
 657 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 658 {
 659         struct nfattr *tb[CTA_HELP_MAX];
 660
 661         DEBUGP("entered %s\n", __FUNCTION__);
 662
 663         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 664
 665         if (!tb[CTA_HELP_NAME-1])
 666                 return -EINVAL;
 667
 668         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 669
 670         return 0;
 671 }
 672
 673 static int
 674 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 675                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 676 {
 677         struct ip_conntrack_tuple_hash *h;
 678         struct ip_conntrack_tuple tuple;
 679         struct ip_conntrack *ct;
 680         int err = 0;
 681
 682         DEBUGP("entered %s\n", __FUNCTION__);
 683
 684         if (cda[CTA_TUPLE_ORIG-1])
 685                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 686         else if (cda[CTA_TUPLE_REPLY-1])
 687                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 688         else {
 689                 /* Flush the whole table */
 690                 ip_conntrack_flush();
 691                 return 0;
 692         }
 693
 694         if (err < 0)
 695                 return err;
 696
 697         h = ip_conntrack_find_get(&tuple, NULL);
 698         if (!h) {
 699                 DEBUGP("tuple not found in conntrack hash\n");
 700                 return -ENOENT;
 701         }
 702
 703         ct = tuplehash_to_ctrack(h);
 704
 705         if (cda[CTA_ID-1]) {
 706                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 707                 if (ct->id != id) {
 708                         ip_conntrack_put(ct);
 709                         return -ENOENT;
 710                 }
 711         }
 712         if (del_timer(&ct->timeout)) {
 713                 ip_conntrack_put(ct);
 714                 ct->timeout.function((unsigned long)ct);
 715                 return 0;
 716         }
 717         ip_conntrack_put(ct);
 718         DEBUGP("leaving\n");
 719
 720         return 0;
 721 }
 722
 723 static int
 724 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 725                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 726 {
 727         struct ip_conntrack_tuple_hash *h;
 728         struct ip_conntrack_tuple tuple;
 729         struct ip_conntrack *ct;
 730         struct sk_buff *skb2 = NULL;
 731         int err = 0;
 732
 733         DEBUGP("entered %s\n", __FUNCTION__);
 734
 735         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 736                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 737                 u32 rlen;
 738
 739                 if (msg->nfgen_family != AF_INET)
 740                         return -EAFNOSUPPORT;
 741
 742                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 743                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 744 #ifdef CONFIG_IP_NF_CT_ACCT
 745                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 746                                                 ctnetlink_dump_table_w,
 747                                                 ctnetlink_done)) != 0)
 748                                 return -EINVAL;
 749 #else
 750                         return -ENOTSUPP;
 751 #endif
 752                 } else {
 753                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 754                                                         ctnetlink_dump_table,
 755                                                         ctnetlink_done)) != 0)
 756                         return -EINVAL;
 757                 }
 758
 759                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 760                 if (rlen > skb->len)
 761                         rlen = skb->len;
 762                 skb_pull(skb, rlen);
 763                 return 0;
 764         }
 765
 766         if (cda[CTA_TUPLE_ORIG-1])
 767                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 768         else if (cda[CTA_TUPLE_REPLY-1])
 769                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 770         else
 771                 return -EINVAL;
 772
 773         if (err < 0)
 774                 return err;
 775
 776         h = ip_conntrack_find_get(&tuple, NULL);
 777         if (!h) {
 778                 DEBUGP("tuple not found in conntrack hash");
 779                 return -ENOENT;
 780         }
 781         DEBUGP("tuple found\n");
 782         ct = tuplehash_to_ctrack(h);
 783
 784         err = -ENOMEM;
 785         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 786         if (!skb2) {
 787                 ip_conntrack_put(ct);
 788                 return -ENOMEM;
 789         }
 790         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 791
 792         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 793                                   IPCTNL_MSG_CT_NEW, 1, ct);
 794         ip_conntrack_put(ct);
 795         if (err <= 0)
 796                 goto free;
 797
 798         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 799         if (err < 0)
 800                 goto out;
 801
 802         DEBUGP("leaving\n");
 803         return 0;
 804
 805 free:
 806         kfree_skb(skb2);
 807 out:
 808         return -1;
 809 }
 810
 811 static inline int
 812 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 813 {
 814         unsigned long d;
 815         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 816         d = ct->status ^ status;
 817
 818         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 819                 /* unchangeable */
 820                 return -EINVAL;
 821
 822         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 823                 /* SEEN_REPLY bit can only be set */
 824                 return -EINVAL;
 825
 826
 827         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 828                 /* ASSURED bit can only be set */
 829                 return -EINVAL;
 830
 831         if (cda[CTA_NAT-1]) {
 832 #ifndef CONFIG_IP_NF_NAT_NEEDED
 833                 return -EINVAL;
 834 #else
 835                 unsigned int hooknum;
 836                 struct ip_nat_range range;
 837
 838                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 839                         return -EINVAL;
 840
 841                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 842                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 843                        htons(range.min.all), htons(range.max.all));
 844
 845                 /* This is tricky but it works. ip_nat_setup_info needs the
 846                  * hook number as parameter, so let's do the correct
 847                  * conversion and run away */
 848                 if (status & IPS_SRC_NAT_DONE)
 849                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 850                 else if (status & IPS_DST_NAT_DONE)
 851                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 852                 else
 853                         return -EINVAL; /* Missing NAT flags */
 854
 855                 DEBUGP("NAT status: %lu\n",
 856                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 857
 858                 if (ip_nat_initialized(ct, hooknum))
 859                         return -EEXIST;
 860                 ip_nat_setup_info(ct, &range, hooknum);
 861
 862                 DEBUGP("NAT status after setup_info: %lu\n",
 863                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 864 #endif
 865         }
 866
 867         /* Be careful here, modifying NAT bits can screw up things,
 868          * so don't let users modify them directly if they don't pass
 869          * ip_nat_range. */
 870         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 871         return 0;
 872 }
 873
 874
 875 static inline int
 876 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 877 {
 878         struct ip_conntrack_helper *helper;
 879         char *helpname;
 880         int err;
 881
 882         DEBUGP("entered %s\n", __FUNCTION__);
 883
 884         /* don't change helper of sibling connections */
 885         if (ct->master)
 886                 return -EINVAL;
 887
 888         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 889         if (err < 0)
 890                 return err;
 891
 892         helper = __ip_conntrack_helper_find_byname(helpname);
 893         if (!helper) {
 894                 if (!strcmp(helpname, ""))
 895                         helper = NULL;
 896                 else
 897                         return -EINVAL;
 898         }
 899
 900         if (ct->helper) {
 901                 if (!helper) {
 902                         /* we had a helper before ... */
 903                         ip_ct_remove_expectations(ct);
 904                         ct->helper = NULL;
 905                 } else {
 906                         /* need to zero data of old helper */
 907                         memset(&ct->help, 0, sizeof(ct->help));
 908                 }
 909         }
 910
 911         ct->helper = helper;
 912
 913         return 0;
 914 }
 915
 916 static inline int
 917 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 918 {
 919         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 920
 921         if (!del_timer(&ct->timeout))
 922                 return -ETIME;
 923
 924         ct->timeout.expires = jiffies + timeout * HZ;
 925         add_timer(&ct->timeout);
 926
 927         return 0;
 928 }
 929
 930 static inline int
 931 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 932 {
 933         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 934         struct ip_conntrack_protocol *proto;
 935         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 936         int err = 0;
 937
 938         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 939
 940         proto = ip_conntrack_proto_find_get(npt);
 941         if (!proto)
 942                 return -EINVAL;
 943
 944         if (proto->from_nfattr)
 945                 err = proto->from_nfattr(tb, ct);
 946         ip_conntrack_proto_put(proto);
 947
 948         return err;
 949 }
 950
 951 static int
 952 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 953 {
 954         int err;
 955
 956         DEBUGP("entered %s\n", __FUNCTION__);
 957
 958         if (cda[CTA_HELP-1]) {
 959                 err = ctnetlink_change_helper(ct, cda);
 960                 if (err < 0)
 961                         return err;
 962         }
 963
 964         if (cda[CTA_TIMEOUT-1]) {
 965                 err = ctnetlink_change_timeout(ct, cda);
 966                 if (err < 0)
 967                         return err;
 968         }
 969
 970         if (cda[CTA_STATUS-1]) {
 971                 err = ctnetlink_change_status(ct, cda);
 972                 if (err < 0)
 973                         return err;
 974         }
 975
 976         if (cda[CTA_PROTOINFO-1]) {
 977                 err = ctnetlink_change_protoinfo(ct, cda);
 978                 if (err < 0)
 979                         return err;
 980         }
 981
 982 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 983         if (cda[CTA_MARK-1])
 984                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
 985 #endif
 986
 987         DEBUGP("all done\n");
 988         return 0;
 989 }
 990
 991 static int
 992 ctnetlink_create_conntrack(struct nfattr *cda[],
 993                            struct ip_conntrack_tuple *otuple,
 994                            struct ip_conntrack_tuple *rtuple)
 995 {
 996         struct ip_conntrack *ct;
 997         int err = -EINVAL;
 998
 999         DEBUGP("entered %s\n", __FUNCTION__);
1000
1001         ct = ip_conntrack_alloc(otuple, rtuple);
1002         if (ct == NULL || IS_ERR(ct))
1003                 return -ENOMEM;
1004
1005         if (!cda[CTA_TIMEOUT-1])
1006                 goto err;
1007         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1008
1009         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1010         ct->status |= IPS_CONFIRMED;
1011
1012         err = ctnetlink_change_status(ct, cda);
1013         if (err < 0)
1014                 goto err;
1015
1016         if (cda[CTA_PROTOINFO-1]) {
1017                 err = ctnetlink_change_protoinfo(ct, cda);
1018                 if (err < 0)
1019                         return err;
1020         }
1021
1022         ct->helper = ip_conntrack_helper_find_get(rtuple);
1023
1024         add_timer(&ct->timeout);
1025         ip_conntrack_hash_insert(ct);
1026
1027         if (ct->helper)
1028                 ip_conntrack_helper_put(ct->helper);
1029
1030 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1031         if (cda[CTA_MARK-1])
1032                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1033 #endif
1034
1035         DEBUGP("conntrack with id %u inserted\n", ct->id);
1036         return 0;
1037
1038 err:
1039         ip_conntrack_free(ct);
1040         return err;
1041 }
1042
1043 static int
1044 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1045                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1046 {
1047         struct ip_conntrack_tuple otuple, rtuple;
1048         struct ip_conntrack_tuple_hash *h = NULL;
1049         int err = 0;
1050
1051         DEBUGP("entered %s\n", __FUNCTION__);
1052
1053         if (cda[CTA_TUPLE_ORIG-1]) {
1054                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1055                 if (err < 0)
1056                         return err;
1057         }
1058
1059         if (cda[CTA_TUPLE_REPLY-1]) {
1060                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1061                 if (err < 0)
1062                         return err;
1063         }
1064
1065         write_lock_bh(&ip_conntrack_lock);
1066         if (cda[CTA_TUPLE_ORIG-1])
1067                 h = __ip_conntrack_find(&otuple, NULL);
1068         else if (cda[CTA_TUPLE_REPLY-1])
1069                 h = __ip_conntrack_find(&rtuple, NULL);
1070
1071         if (h == NULL) {
1072                 write_unlock_bh(&ip_conntrack_lock);
1073                 DEBUGP("no such conntrack, create new\n");
1074                 err = -ENOENT;
1075                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1076                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1077                 return err;
1078         }
1079         /* implicit 'else' */
1080
1081         /* we only allow nat config for new conntracks */
1082         if (cda[CTA_NAT-1]) {
1083                 err = -EINVAL;
1084                 goto out_unlock;
1085         }
1086
1087         /* We manipulate the conntrack inside the global conntrack table lock,
1088          * so there's no need to increase the refcount */
1089         DEBUGP("conntrack found\n");
1090         err = -EEXIST;
1091         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1092                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1093
1094 out_unlock:
1095         write_unlock_bh(&ip_conntrack_lock);
1096         return err;
1097 }
1098
1099 /***********************************************************************
1100  * EXPECT
1101  ***********************************************************************/
1102
1103 static inline int
1104 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1105                          const struct ip_conntrack_tuple *tuple,
1106                          enum ctattr_expect type)
1107 {
1108         struct nfattr *nest_parms = NFA_NEST(skb, type);
1109
1110         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1111                 goto nfattr_failure;
1112
1113         NFA_NEST_END(skb, nest_parms);
1114
1115         return 0;
1116
1117 nfattr_failure:
1118         return -1;
1119 }
1120
1121 static inline int
1122 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1123                           const struct ip_conntrack_expect *exp)
1124 {
1125         struct ip_conntrack *master = exp->master;
1126         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1127         u_int32_t id = htonl(exp->id);
1128
1129         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1130                 goto nfattr_failure;
1131         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1132                 goto nfattr_failure;
1133         if (ctnetlink_exp_dump_tuple(skb,
1134                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1135                                  CTA_EXPECT_MASTER) < 0)
1136                 goto nfattr_failure;
1137
1138         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1139         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1140
1141         return 0;
1142
1143 nfattr_failure:
1144         return -1;
1145 }
1146
1147 static int
1148 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1149                     int event,
1150                     int nowait,
1151                     const struct ip_conntrack_expect *exp)
1152 {
1153         struct nlmsghdr *nlh;
1154         struct nfgenmsg *nfmsg;
1155         unsigned char *b;
1156
1157         b = skb->tail;
1158
1159         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1160         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1161         nfmsg  = NLMSG_DATA(nlh);
1162
1163         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1164         nfmsg->nfgen_family = AF_INET;
1165         nfmsg->version      = NFNETLINK_V0;
1166         nfmsg->res_id       = 0;
1167
1168         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1169                 goto nfattr_failure;
1170
1171         nlh->nlmsg_len = skb->tail - b;
1172         return skb->len;
1173
1174 nlmsg_failure:
1175 nfattr_failure:
1176         skb_trim(skb, b - skb->data);
1177         return -1;
1178 }
1179
1180 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1181 static int ctnetlink_expect_event(struct notifier_block *this,
1182                                   unsigned long events, void *ptr)
1183 {
1184         struct nlmsghdr *nlh;
1185         struct nfgenmsg *nfmsg;
1186         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1187         struct sk_buff *skb;
1188         unsigned int type;
1189         unsigned char *b;
1190         int flags = 0;
1191         u16 proto;
1192
1193         if (events & IPEXP_NEW) {
1194                 type = IPCTNL_MSG_EXP_NEW;
1195                 flags = NLM_F_CREATE|NLM_F_EXCL;
1196         } else
1197                 return NOTIFY_DONE;
1198
1199         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1200         if (!skb)
1201                 return NOTIFY_DONE;
1202
1203         b = skb->tail;
1204
1205         type |= NFNL_SUBSYS_CTNETLINK << 8;
1206         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1207         nfmsg = NLMSG_DATA(nlh);
1208
1209         nlh->nlmsg_flags    = flags;
1210         nfmsg->nfgen_family = AF_INET;
1211         nfmsg->version      = NFNETLINK_V0;
1212         nfmsg->res_id       = 0;
1213
1214         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1215                 goto nfattr_failure;
1216
1217         nlh->nlmsg_len = skb->tail - b;
1218         proto = exp->tuple.dst.protonum;
1219         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1220         return NOTIFY_DONE;
1221
1222 nlmsg_failure:
1223 nfattr_failure:
1224         kfree_skb(skb);
1225         return NOTIFY_DONE;
1226 }
1227 #endif
1228
1229 static int
1230 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1231 {
1232         struct ip_conntrack_expect *exp = NULL;
1233         struct list_head *i;
1234         u_int32_t *id = (u_int32_t *) &cb->args[0];
1235
1236         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1237
1238         read_lock_bh(&ip_conntrack_lock);
1239         list_for_each_prev(i, &ip_conntrack_expect_list) {
1240                 exp = (struct ip_conntrack_expect *) i;
1241                 if (exp->id <= *id)
1242                         continue;
1243                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1244                                             cb->nlh->nlmsg_seq,
1245                                             IPCTNL_MSG_EXP_NEW,
1246                                             1, exp) < 0)
1247                         goto out;
1248                 *id = exp->id;
1249         }
1250 out:
1251         read_unlock_bh(&ip_conntrack_lock);
1252
1253         DEBUGP("leaving, last id=%llu\n", *id);
1254
1255         return skb->len;
1256 }
1257
1258 static int
1259 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1260                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1261 {
1262         struct ip_conntrack_tuple tuple;
1263         struct ip_conntrack_expect *exp;
1264         struct sk_buff *skb2;
1265         int err = 0;
1266
1267         DEBUGP("entered %s\n", __FUNCTION__);
1268
1269         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1270                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1271                 u32 rlen;
1272
1273                 if (msg->nfgen_family != AF_INET)
1274                         return -EAFNOSUPPORT;
1275
1276                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1277                                                 ctnetlink_exp_dump_table,
1278                                                 ctnetlink_done)) != 0)
1279                         return -EINVAL;
1280                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1281                 if (rlen > skb->len)
1282                         rlen = skb->len;
1283                 skb_pull(skb, rlen);
1284                 return 0;
1285         }
1286
1287         if (cda[CTA_EXPECT_MASTER-1])
1288                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1289         else
1290                 return -EINVAL;
1291
1292         if (err < 0)
1293                 return err;
1294
1295         exp = ip_conntrack_expect_find(&tuple);
1296         if (!exp)
1297                 return -ENOENT;
1298
1299         err = -ENOMEM;
1300         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1301         if (!skb2)
1302                 goto out;
1303         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1304
1305         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1306                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1307                                       1, exp);
1308         if (err <= 0)
1309                 goto free;
1310
1311         ip_conntrack_expect_put(exp);
1312
1313         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1314
1315 free:
1316         kfree_skb(skb2);
1317 out:
1318         ip_conntrack_expect_put(exp);
1319         return err;
1320 }
1321
1322 static int
1323 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1324                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1325 {
1326         struct ip_conntrack_expect *exp, *tmp;
1327         struct ip_conntrack_tuple tuple;
1328         struct ip_conntrack_helper *h;
1329         int err;
1330
1331         if (cda[CTA_EXPECT_TUPLE-1]) {
1332                 /* delete a single expect by tuple */
1333                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1334                 if (err < 0)
1335                         return err;
1336
1337                 /* bump usage count to 2 */
1338                 exp = ip_conntrack_expect_find(&tuple);
1339                 if (!exp)
1340                         return -ENOENT;
1341
1342                 if (cda[CTA_EXPECT_ID-1]) {
1343                         u_int32_t id =
1344                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1345                         if (exp->id != ntohl(id)) {
1346                                 ip_conntrack_expect_put(exp);
1347                                 return -ENOENT;
1348                         }
1349                 }
1350
1351                 /* after list removal, usage count == 1 */
1352                 ip_conntrack_unexpect_related(exp);
1353                 /* have to put what we 'get' above.
1354                  * after this line usage count == 0 */
1355                 ip_conntrack_expect_put(exp);
1356         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1357                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1358
1359                 /* delete all expectations for this helper */
1360                 write_lock_bh(&ip_conntrack_lock);
1361                 h = __ip_conntrack_helper_find_byname(name);
1362                 if (!h) {
1363                         write_unlock_bh(&ip_conntrack_lock);
1364                         return -EINVAL;
1365                 }
1366                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1367                                          list) {
1368                         if (exp->master->helper == h
1369                             && del_timer(&exp->timeout)) {
1370                                 ip_ct_unlink_expect(exp);
1371                                 ip_conntrack_expect_put(exp);
1372                         }
1373                 }
1374                 write_unlock(&ip_conntrack_lock);
1375         } else {
1376                 /* This basically means we have to flush everything*/
1377                 write_lock_bh(&ip_conntrack_lock);
1378                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1379                                          list) {
1380                         if (del_timer(&exp->timeout)) {
1381                                 ip_ct_unlink_expect(exp);
1382                                 ip_conntrack_expect_put(exp);
1383                         }
1384                 }
1385                 write_unlock_bh(&ip_conntrack_lock);
1386         }
1387
1388         return 0;
1389 }
1390 static int
1391 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1392 {
1393         return -EOPNOTSUPP;
1394 }
1395
1396 static int
1397 ctnetlink_create_expect(struct nfattr *cda[])
1398 {
1399         struct ip_conntrack_tuple tuple, mask, master_tuple;
1400         struct ip_conntrack_tuple_hash *h = NULL;
1401         struct ip_conntrack_expect *exp;
1402         struct ip_conntrack *ct;
1403         int err = 0;
1404
1405         DEBUGP("entered %s\n", __FUNCTION__);
1406
1407         /* caller guarantees that those three CTA_EXPECT_* exist */
1408         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1409         if (err < 0)
1410                 return err;
1411         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1412         if (err < 0)
1413                 return err;
1414         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1415         if (err < 0)
1416                 return err;
1417
1418         /* Look for master conntrack of this expectation */
1419         h = ip_conntrack_find_get(&master_tuple, NULL);
1420         if (!h)
1421                 return -ENOENT;
1422         ct = tuplehash_to_ctrack(h);
1423
1424         if (!ct->helper) {
1425                 /* such conntrack hasn't got any helper, abort */
1426                 err = -EINVAL;
1427                 goto out;
1428         }
1429
1430         exp = ip_conntrack_expect_alloc(ct);
1431         if (!exp) {
1432                 err = -ENOMEM;
1433                 goto out;
1434         }
1435
1436         exp->expectfn = NULL;
1437         exp->flags = 0;
1438         exp->master = ct;
1439         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1440         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1441
1442         err = ip_conntrack_expect_related(exp);
1443         ip_conntrack_expect_put(exp);
1444
1445 out:
1446         ip_conntrack_put(tuplehash_to_ctrack(h));
1447         return err;
1448 }
1449
1450 static int
1451 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1452                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1453 {
1454         struct ip_conntrack_tuple tuple;
1455         struct ip_conntrack_expect *exp;
1456         int err = 0;
1457
1458         DEBUGP("entered %s\n", __FUNCTION__);
1459
1460         if (!cda[CTA_EXPECT_TUPLE-1]
1461             || !cda[CTA_EXPECT_MASK-1]
1462             || !cda[CTA_EXPECT_MASTER-1])
1463                 return -EINVAL;
1464
1465         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1466         if (err < 0)
1467                 return err;
1468
1469         write_lock_bh(&ip_conntrack_lock);
1470         exp = __ip_conntrack_expect_find(&tuple);
1471
1472         if (!exp) {
1473                 write_unlock_bh(&ip_conntrack_lock);
1474                 err = -ENOENT;
1475                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1476                         err = ctnetlink_create_expect(cda);
1477                 return err;
1478         }
1479
1480         err = -EEXIST;
1481         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1482                 err = ctnetlink_change_expect(exp, cda);
1483         write_unlock_bh(&ip_conntrack_lock);
1484
1485         DEBUGP("leaving\n");
1486
1487         return err;
1488 }
1489
1490 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1491 static struct notifier_block ctnl_notifier = {
1492         .notifier_call  = ctnetlink_conntrack_event,
1493 };
1494
1495 static struct notifier_block ctnl_notifier_exp = {
1496         .notifier_call  = ctnetlink_expect_event,
1497 };
1498 #endif
1499
1500 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1501         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1502                                             .attr_count = CTA_MAX,
1503                                             .cap_required = CAP_NET_ADMIN },
1504         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1505                                             .attr_count = CTA_MAX,
1506                                             .cap_required = CAP_NET_ADMIN },
1507         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1508                                             .attr_count = CTA_MAX,
1509                                             .cap_required = CAP_NET_ADMIN },
1510         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1511                                             .attr_count = CTA_MAX,
1512                                             .cap_required = CAP_NET_ADMIN },
1513 };
1514
1515 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1516         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1517                                             .attr_count = CTA_EXPECT_MAX,
1518                                             .cap_required = CAP_NET_ADMIN },
1519         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1520                                             .attr_count = CTA_EXPECT_MAX,
1521                                             .cap_required = CAP_NET_ADMIN },
1522         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1523                                             .attr_count = CTA_EXPECT_MAX,
1524                                             .cap_required = CAP_NET_ADMIN },
1525 };
1526
1527 static struct nfnetlink_subsystem ctnl_subsys = {
1528         .name                           = "conntrack",
1529         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1530         .cb_count                       = IPCTNL_MSG_MAX,
1531         .cb                             = ctnl_cb,
1532 };
1533
1534 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1535         .name                           = "conntrack_expect",
1536         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1537         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1538         .cb                             = ctnl_exp_cb,
1539 };
1540
1541 static int __init ctnetlink_init(void)
1542 {
1543         int ret;
1544
1545         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1546         ret = nfnetlink_subsys_register(&ctnl_subsys);
1547         if (ret < 0) {
1548                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1549                 goto err_out;
1550         }
1551
1552         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1553         if (ret < 0) {
1554                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1555                 goto err_unreg_subsys;
1556         }
1557
1558 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1559         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1560         if (ret < 0) {
1561                 printk("ctnetlink_init: cannot register notifier.\n");
1562                 goto err_unreg_exp_subsys;
1563         }
1564
1565         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1566         if (ret < 0) {
1567                 printk("ctnetlink_init: cannot expect register notifier.\n");
1568                 goto err_unreg_notifier;
1569         }
1570 #endif
1571
1572         return 0;
1573
1574 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1575 err_unreg_notifier:
1576         ip_conntrack_unregister_notifier(&ctnl_notifier);
1577 err_unreg_exp_subsys:
1578         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1579 #endif
1580 err_unreg_subsys:
1581         nfnetlink_subsys_unregister(&ctnl_subsys);
1582 err_out:
1583         return ret;
1584 }
1585
1586 static void __exit ctnetlink_exit(void)
1587 {
1588         printk("ctnetlink: unregistering from nfnetlink.\n");
1589
1590 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1591         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1592         ip_conntrack_unregister_notifier(&ctnl_notifier);
1593 #endif
1594
1595         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1596         nfnetlink_subsys_unregister(&ctnl_subsys);
1597         return;
1598 }
1599
1600 module_init(ctnetlink_init);
1601 module_exit(ctnetlink_exit);