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
  62         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  63
  64         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  65         if (proto && proto->tuple_to_nfattr)
  66                 return proto->tuple_to_nfattr(skb, tuple);
  67
  68         return 0;
  69
  70 nfattr_failure:
  71         return -1;
  72 }
  73
  74 static inline int
  75 ctnetlink_dump_tuples(struct sk_buff *skb,
  76                       const struct ip_conntrack_tuple *tuple)
  77 {
  78         struct nfattr *nest_parms;
  79
  80         nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  81         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  82         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  83         NFA_NEST_END(skb, nest_parms);
  84
  85         nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  86         ctnetlink_dump_tuples_proto(skb, tuple);
  87         NFA_NEST_END(skb, nest_parms);
  88
  89         return 0;
  90
  91 nfattr_failure:
  92         return -1;
  93 }
  94
  95 static inline int
  96 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
  97 {
  98         u_int32_t status = htonl((u_int32_t) ct->status);
  99         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 100         return 0;
 101
 102 nfattr_failure:
 103         return -1;
 104 }
 105
 106 static inline int
 107 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 108 {
 109         long timeout_l = ct->timeout.expires - jiffies;
 110         u_int32_t timeout;
 111
 112         if (timeout_l < 0)
 113                 timeout = 0;
 114         else
 115                 timeout = htonl(timeout_l / HZ);
 116
 117         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 118         return 0;
 119
 120 nfattr_failure:
 121         return -1;
 122 }
 123
 124 static inline int
 125 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 126 {
 127         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 128
 129         struct nfattr *nest_proto;
 130         int ret;
 131
 132         if (!proto || !proto->to_nfattr)
 133                 return 0;
 134
 135         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 136
 137         ret = proto->to_nfattr(skb, nest_proto, ct);
 138
 139         ip_conntrack_proto_put(proto);
 140
 141         NFA_NEST_END(skb, nest_proto);
 142
 143         return ret;
 144
 145 nfattr_failure:
 146         return -1;
 147 }
 148
 149 static inline int
 150 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 151 {
 152         struct nfattr *nest_helper;
 153
 154         if (!ct->helper)
 155                 return 0;
 156
 157         nest_helper = NFA_NEST(skb, CTA_HELP);
 158         NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
 159
 160         if (ct->helper->to_nfattr)
 161                 ct->helper->to_nfattr(skb, ct);
 162
 163         NFA_NEST_END(skb, nest_helper);
 164
 165         return 0;
 166
 167 nfattr_failure:
 168         return -1;
 169 }
 170
 171 #ifdef CONFIG_IP_NF_CT_ACCT
 172 static inline int
 173 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 174                         enum ip_conntrack_dir dir)
 175 {
 176         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 177         struct nfattr *nest_count = NFA_NEST(skb, type);
 178         u_int64_t tmp;
 179
 180         tmp = htonl(ct->counters[dir].packets);
 181         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 182
 183         tmp = htonl(ct->counters[dir].bytes);
 184         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 185
 186         NFA_NEST_END(skb, nest_count);
 187
 188         return 0;
 189
 190 nfattr_failure:
 191         return -1;
 192 }
 193 #else
 194 #define ctnetlink_dump_counters(a, b, c) (0)
 195 #endif
 196
 197 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 198 static inline int
 199 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 200 {
 201         u_int32_t mark = htonl(ct->mark);
 202
 203         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 204         return 0;
 205
 206 nfattr_failure:
 207         return -1;
 208 }
 209 #else
 210 #define ctnetlink_dump_mark(a, b) (0)
 211 #endif
 212
 213 static inline int
 214 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 215 {
 216         u_int32_t id = htonl(ct->id);
 217         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 218         return 0;
 219
 220 nfattr_failure:
 221         return -1;
 222 }
 223
 224 static inline int
 225 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 226 {
 227         unsigned int use = htonl(atomic_read(&ct->ct_general.use));
 228
 229         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 230         return 0;
 231
 232 nfattr_failure:
 233         return -1;
 234 }
 235
 236 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 237
 238 static int
 239 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 240                     int event, int nowait,
 241                     const struct ip_conntrack *ct)
 242 {
 243         struct nlmsghdr *nlh;
 244         struct nfgenmsg *nfmsg;
 245         struct nfattr *nest_parms;
 246         unsigned char *b;
 247
 248         b = skb->tail;
 249
 250         event |= NFNL_SUBSYS_CTNETLINK << 8;
 251         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 252         nfmsg  = NLMSG_DATA(nlh);
 253
 254         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 255         nfmsg->nfgen_family = AF_INET;
 256         nfmsg->version      = NFNETLINK_V0;
 257         nfmsg->res_id       = 0;
 258
 259         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 260         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 261                 goto nfattr_failure;
 262         NFA_NEST_END(skb, nest_parms);
 263
 264         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 265         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 266                 goto nfattr_failure;
 267         NFA_NEST_END(skb, nest_parms);
 268
 269         if (ctnetlink_dump_status(skb, ct) < 0 ||
 270             ctnetlink_dump_timeout(skb, ct) < 0 ||
 271             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 272             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 273             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 274             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 275             ctnetlink_dump_mark(skb, ct) < 0 ||
 276             ctnetlink_dump_id(skb, ct) < 0 ||
 277             ctnetlink_dump_use(skb, ct) < 0)
 278                 goto nfattr_failure;
 279
 280         nlh->nlmsg_len = skb->tail - b;
 281         return skb->len;
 282
 283 nlmsg_failure:
 284 nfattr_failure:
 285         skb_trim(skb, b - skb->data);
 286         return -1;
 287 }
 288
 289 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 290 static int ctnetlink_conntrack_event(struct notifier_block *this,
 291                                      unsigned long events, void *ptr)
 292 {
 293         struct nlmsghdr *nlh;
 294         struct nfgenmsg *nfmsg;
 295         struct nfattr *nest_parms;
 296         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 297         struct sk_buff *skb;
 298         unsigned int type;
 299         unsigned char *b;
 300         unsigned int flags = 0, group;
 301
 302         /* ignore our fake conntrack entry */
 303         if (ct == &ip_conntrack_untracked)
 304                 return NOTIFY_DONE;
 305
 306         if (events & IPCT_DESTROY) {
 307                 type = IPCTNL_MSG_CT_DELETE;
 308                 group = NFNLGRP_CONNTRACK_DESTROY;
 309                 goto alloc_skb;
 310         }
 311         if (events & (IPCT_NEW | IPCT_RELATED)) {
 312                 type = IPCTNL_MSG_CT_NEW;
 313                 flags = NLM_F_CREATE|NLM_F_EXCL;
 314                 /* dump everything */
 315                 events = ~0UL;
 316                 group = NFNLGRP_CONNTRACK_NEW;
 317                 goto alloc_skb;
 318         }
 319         if (events & (IPCT_STATUS |
 320                       IPCT_PROTOINFO |
 321                       IPCT_HELPER |
 322                       IPCT_HELPINFO |
 323                       IPCT_NATINFO)) {
 324                 type = IPCTNL_MSG_CT_NEW;
 325                 group = NFNLGRP_CONNTRACK_UPDATE;
 326                 goto alloc_skb;
 327         }
 328
 329         return NOTIFY_DONE;
 330
 331 alloc_skb:
 332   /* FIXME: Check if there are any listeners before, don't hurt performance */
 333
 334         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 335         if (!skb)
 336                 return NOTIFY_DONE;
 337
 338         b = skb->tail;
 339
 340         type |= NFNL_SUBSYS_CTNETLINK << 8;
 341         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 342         nfmsg = NLMSG_DATA(nlh);
 343
 344         nlh->nlmsg_flags    = flags;
 345         nfmsg->nfgen_family = AF_INET;
 346         nfmsg->version  = NFNETLINK_V0;
 347         nfmsg->res_id   = 0;
 348
 349         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 350         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 351                 goto nfattr_failure;
 352         NFA_NEST_END(skb, nest_parms);
 353
 354         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 355         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 356                 goto nfattr_failure;
 357         NFA_NEST_END(skb, nest_parms);
 358
 359         /* NAT stuff is now a status flag */
 360         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 361             && ctnetlink_dump_status(skb, ct) < 0)
 362                 goto nfattr_failure;
 363         if (events & IPCT_REFRESH
 364             && ctnetlink_dump_timeout(skb, ct) < 0)
 365                 goto nfattr_failure;
 366         if (events & IPCT_PROTOINFO
 367             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 368                 goto nfattr_failure;
 369         if (events & IPCT_HELPINFO
 370             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 371                 goto nfattr_failure;
 372
 373         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 374             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 375                 goto nfattr_failure;
 376
 377         nlh->nlmsg_len = skb->tail - b;
 378         nfnetlink_send(skb, 0, group, 0);
 379         return NOTIFY_DONE;
 380
 381 nlmsg_failure:
 382 nfattr_failure:
 383         kfree_skb(skb);
 384         return NOTIFY_DONE;
 385 }
 386 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 387
 388 static int ctnetlink_done(struct netlink_callback *cb)
 389 {
 390         DEBUGP("entered %s\n", __FUNCTION__);
 391         return 0;
 392 }
 393
 394 static int
 395 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 396 {
 397         struct ip_conntrack *ct = NULL;
 398         struct ip_conntrack_tuple_hash *h;
 399         struct list_head *i;
 400         u_int32_t *id = (u_int32_t *) &cb->args[1];
 401
 402         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 403                         cb->args[0], *id);
 404
 405         read_lock_bh(&ip_conntrack_lock);
 406         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 407                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 408                         h = (struct ip_conntrack_tuple_hash *) i;
 409                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 410                                 continue;
 411                         ct = tuplehash_to_ctrack(h);
 412                         if (ct->id <= *id)
 413                                 continue;
 414                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 415                                                 cb->nlh->nlmsg_seq,
 416                                                 IPCTNL_MSG_CT_NEW,
 417                                                 1, ct) < 0)
 418                                 goto out;
 419                         *id = ct->id;
 420                 }
 421         }
 422 out:
 423         read_unlock_bh(&ip_conntrack_lock);
 424
 425         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 426
 427         return skb->len;
 428 }
 429
 430 #ifdef CONFIG_IP_NF_CT_ACCT
 431 static int
 432 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 433 {
 434         struct ip_conntrack *ct = NULL;
 435         struct ip_conntrack_tuple_hash *h;
 436         struct list_head *i;
 437         u_int32_t *id = (u_int32_t *) &cb->args[1];
 438
 439         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 440                         cb->args[0], *id);
 441
 442         write_lock_bh(&ip_conntrack_lock);
 443         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 444                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 445                         h = (struct ip_conntrack_tuple_hash *) i;
 446                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 447                                 continue;
 448                         ct = tuplehash_to_ctrack(h);
 449                         if (ct->id <= *id)
 450                                 continue;
 451                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 452                                                 cb->nlh->nlmsg_seq,
 453                                                 IPCTNL_MSG_CT_NEW,
 454                                                 1, ct) < 0)
 455                                 goto out;
 456                         *id = ct->id;
 457
 458                         memset(&ct->counters, 0, sizeof(ct->counters));
 459                 }
 460         }
 461 out:
 462         write_unlock_bh(&ip_conntrack_lock);
 463
 464         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 465
 466         return skb->len;
 467 }
 468 #endif
 469
 470 static const int cta_min_ip[CTA_IP_MAX] = {
 471         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 472         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 473 };
 474
 475 static inline int
 476 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 477 {
 478         struct nfattr *tb[CTA_IP_MAX];
 479
 480         DEBUGP("entered %s\n", __FUNCTION__);
 481
 482
 483         if (nfattr_parse_nested(tb, CTA_IP_MAX, attr) < 0)
 484                 goto nfattr_failure;
 485
 486         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 487                 return -EINVAL;
 488
 489         if (!tb[CTA_IP_V4_SRC-1])
 490                 return -EINVAL;
 491         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 492
 493         if (!tb[CTA_IP_V4_DST-1])
 494                 return -EINVAL;
 495         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 496
 497         DEBUGP("leaving\n");
 498
 499         return 0;
 500
 501 nfattr_failure:
 502         return -1;
 503 }
 504
 505 static const int 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         if (nfattr_parse_nested(tb, CTA_PROTO_MAX, attr) < 0)
 525                 goto nfattr_failure;
 526
 527         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 528                 return -EINVAL;
 529
 530         if (!tb[CTA_PROTO_NUM-1])
 531                 return -EINVAL;
 532         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 533
 534         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 535
 536         if (likely(proto && proto->nfattr_to_tuple)) {
 537                 ret = proto->nfattr_to_tuple(tb, tuple);
 538                 ip_conntrack_proto_put(proto);
 539         }
 540
 541         return ret;
 542
 543 nfattr_failure:
 544         return -1;
 545 }
 546
 547 static inline int
 548 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 549                       enum ctattr_tuple type)
 550 {
 551         struct nfattr *tb[CTA_TUPLE_MAX];
 552         int err;
 553
 554         DEBUGP("entered %s\n", __FUNCTION__);
 555
 556         memset(tuple, 0, sizeof(*tuple));
 557
 558         if (nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]) < 0)
 559                 goto nfattr_failure;
 560
 561         if (!tb[CTA_TUPLE_IP-1])
 562                 return -EINVAL;
 563
 564         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 565         if (err < 0)
 566                 return err;
 567
 568         if (!tb[CTA_TUPLE_PROTO-1])
 569                 return -EINVAL;
 570
 571         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 572         if (err < 0)
 573                 return err;
 574
 575         /* orig and expect tuples get DIR_ORIGINAL */
 576         if (type == CTA_TUPLE_REPLY)
 577                 tuple->dst.dir = IP_CT_DIR_REPLY;
 578         else
 579                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 580
 581         DUMP_TUPLE(tuple);
 582
 583         DEBUGP("leaving\n");
 584
 585         return 0;
 586
 587 nfattr_failure:
 588         return -1;
 589 }
 590
 591 #ifdef CONFIG_IP_NF_NAT_NEEDED
 592 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
 593         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 594         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 595 };
 596
 597 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 598                                      const struct ip_conntrack *ct,
 599                                      struct ip_nat_range *range)
 600 {
 601         struct nfattr *tb[CTA_PROTONAT_MAX];
 602         struct ip_nat_protocol *npt;
 603
 604         DEBUGP("entered %s\n", __FUNCTION__);
 605
 606         if (nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr) < 0)
 607                 goto nfattr_failure;
 608
 609         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 610                 goto nfattr_failure;
 611
 612         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 613         if (!npt)
 614                 return 0;
 615
 616         if (!npt->nfattr_to_range) {
 617                 ip_nat_proto_put(npt);
 618                 return 0;
 619         }
 620
 621         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 622         if (npt->nfattr_to_range(tb, range) > 0)
 623                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 624
 625         ip_nat_proto_put(npt);
 626
 627         DEBUGP("leaving\n");
 628         return 0;
 629
 630 nfattr_failure:
 631         return -1;
 632 }
 633
 634 static inline int
 635 ctnetlink_parse_nat(struct nfattr *cda[],
 636                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 637 {
 638         struct nfattr *tb[CTA_NAT_MAX];
 639         int err;
 640
 641         DEBUGP("entered %s\n", __FUNCTION__);
 642
 643         memset(range, 0, sizeof(*range));
 644
 645         if (nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]) < 0)
 646                 goto nfattr_failure;
 647
 648         if (tb[CTA_NAT_MINIP-1])
 649                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 650
 651         if (!tb[CTA_NAT_MAXIP-1])
 652                 range->max_ip = range->min_ip;
 653         else
 654                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 655
 656         if (range->min_ip)
 657                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 658
 659         if (!tb[CTA_NAT_PROTO-1])
 660                 return 0;
 661
 662         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 663         if (err < 0)
 664                 return err;
 665
 666         DEBUGP("leaving\n");
 667         return 0;
 668
 669 nfattr_failure:
 670         return -1;
 671 }
 672 #endif
 673
 674 static inline int
 675 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 676 {
 677         struct nfattr *tb[CTA_HELP_MAX];
 678
 679         DEBUGP("entered %s\n", __FUNCTION__);
 680
 681         if (nfattr_parse_nested(tb, CTA_HELP_MAX, attr) < 0)
 682                 goto nfattr_failure;
 683
 684         if (!tb[CTA_HELP_NAME-1])
 685                 return -EINVAL;
 686
 687         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 688
 689         return 0;
 690
 691 nfattr_failure:
 692         return -1;
 693 }
 694
 695 static int
 696 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 697                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 698 {
 699         struct ip_conntrack_tuple_hash *h;
 700         struct ip_conntrack_tuple tuple;
 701         struct ip_conntrack *ct;
 702         int err = 0;
 703
 704         DEBUGP("entered %s\n", __FUNCTION__);
 705
 706         if (cda[CTA_TUPLE_ORIG-1])
 707                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 708         else if (cda[CTA_TUPLE_REPLY-1])
 709                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 710         else {
 711                 /* Flush the whole table */
 712                 ip_conntrack_flush();
 713                 return 0;
 714         }
 715
 716         if (err < 0)
 717                 return err;
 718
 719         h = ip_conntrack_find_get(&tuple, NULL);
 720         if (!h) {
 721                 DEBUGP("tuple not found in conntrack hash\n");
 722                 return -ENOENT;
 723         }
 724
 725         ct = tuplehash_to_ctrack(h);
 726
 727         if (cda[CTA_ID-1]) {
 728                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 729                 if (ct->id != id) {
 730                         ip_conntrack_put(ct);
 731                         return -ENOENT;
 732                 }
 733         }
 734         if (del_timer(&ct->timeout)) {
 735                 ip_conntrack_put(ct);
 736                 ct->timeout.function((unsigned long)ct);
 737                 return 0;
 738         }
 739         ip_conntrack_put(ct);
 740         DEBUGP("leaving\n");
 741
 742         return 0;
 743 }
 744
 745 static int
 746 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 747                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 748 {
 749         struct ip_conntrack_tuple_hash *h;
 750         struct ip_conntrack_tuple tuple;
 751         struct ip_conntrack *ct;
 752         struct sk_buff *skb2 = NULL;
 753         int err = 0;
 754
 755         DEBUGP("entered %s\n", __FUNCTION__);
 756
 757         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 758                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 759                 u32 rlen;
 760
 761                 if (msg->nfgen_family != AF_INET)
 762                         return -EAFNOSUPPORT;
 763
 764                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 765                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 766 #ifdef CONFIG_IP_NF_CT_ACCT
 767                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 768                                                 ctnetlink_dump_table_w,
 769                                                 ctnetlink_done)) != 0)
 770                                 return -EINVAL;
 771 #else
 772                         return -ENOTSUPP;
 773 #endif
 774                 } else {
 775                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 776                                                         ctnetlink_dump_table,
 777                                                         ctnetlink_done)) != 0)
 778                         return -EINVAL;
 779                 }
 780
 781                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 782                 if (rlen > skb->len)
 783                         rlen = skb->len;
 784                 skb_pull(skb, rlen);
 785                 return 0;
 786         }
 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_ATOMIC);
 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 out;
 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 out:
 828         if (skb2)
 829                 kfree_skb(skb2);
 830         return -1;
 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         if (nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr) < 0)
 961                 goto nfattr_failure;
 962
 963         proto = ip_conntrack_proto_find_get(npt);
 964         if (!proto)
 965                 return -EINVAL;
 966
 967         if (proto->from_nfattr)
 968                 err = proto->from_nfattr(tb, ct);
 969         ip_conntrack_proto_put(proto);
 970
 971         return err;
 972
 973 nfattr_failure:
 974         return -ENOMEM;
 975 }
 976
 977 static int
 978 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 979 {
 980         int err;
 981
 982         DEBUGP("entered %s\n", __FUNCTION__);
 983
 984         if (cda[CTA_HELP-1]) {
 985                 err = ctnetlink_change_helper(ct, cda);
 986                 if (err < 0)
 987                         return err;
 988         }
 989
 990         if (cda[CTA_TIMEOUT-1]) {
 991                 err = ctnetlink_change_timeout(ct, cda);
 992                 if (err < 0)
 993                         return err;
 994         }
 995
 996         if (cda[CTA_STATUS-1]) {
 997                 err = ctnetlink_change_status(ct, cda);
 998                 if (err < 0)
 999                         return err;
1000         }
1001
1002         if (cda[CTA_PROTOINFO-1]) {
1003                 err = ctnetlink_change_protoinfo(ct, cda);
1004                 if (err < 0)
1005                         return err;
1006         }
1007
1008         DEBUGP("all done\n");
1009         return 0;
1010 }
1011
1012 static int
1013 ctnetlink_create_conntrack(struct nfattr *cda[],
1014                            struct ip_conntrack_tuple *otuple,
1015                            struct ip_conntrack_tuple *rtuple)
1016 {
1017         struct ip_conntrack *ct;
1018         int err = -EINVAL;
1019
1020         DEBUGP("entered %s\n", __FUNCTION__);
1021
1022         ct = ip_conntrack_alloc(otuple, rtuple);
1023         if (ct == NULL || IS_ERR(ct))
1024                 return -ENOMEM;
1025
1026         if (!cda[CTA_TIMEOUT-1])
1027                 goto err;
1028         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1029
1030         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1031         ct->status |= IPS_CONFIRMED;
1032
1033         err = ctnetlink_change_status(ct, cda);
1034         if (err < 0)
1035                 goto err;
1036
1037         if (cda[CTA_PROTOINFO-1]) {
1038                 err = ctnetlink_change_protoinfo(ct, cda);
1039                 if (err < 0)
1040                         return err;
1041         }
1042
1043         ct->helper = ip_conntrack_helper_find_get(rtuple);
1044
1045         add_timer(&ct->timeout);
1046         ip_conntrack_hash_insert(ct);
1047
1048         if (ct->helper)
1049                 ip_conntrack_helper_put(ct->helper);
1050
1051         DEBUGP("conntrack with id %u inserted\n", ct->id);
1052         return 0;
1053
1054 err:
1055         ip_conntrack_free(ct);
1056         return err;
1057 }
1058
1059 static int
1060 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1061                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1062 {
1063         struct ip_conntrack_tuple otuple, rtuple;
1064         struct ip_conntrack_tuple_hash *h = NULL;
1065         int err = 0;
1066
1067         DEBUGP("entered %s\n", __FUNCTION__);
1068
1069         if (cda[CTA_TUPLE_ORIG-1]) {
1070                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1071                 if (err < 0)
1072                         return err;
1073         }
1074
1075         if (cda[CTA_TUPLE_REPLY-1]) {
1076                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1077                 if (err < 0)
1078                         return err;
1079         }
1080
1081         write_lock_bh(&ip_conntrack_lock);
1082         if (cda[CTA_TUPLE_ORIG-1])
1083                 h = __ip_conntrack_find(&otuple, NULL);
1084         else if (cda[CTA_TUPLE_REPLY-1])
1085                 h = __ip_conntrack_find(&rtuple, NULL);
1086
1087         if (h == NULL) {
1088                 write_unlock_bh(&ip_conntrack_lock);
1089                 DEBUGP("no such conntrack, create new\n");
1090                 err = -ENOENT;
1091                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1092                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1093                 return err;
1094         }
1095         /* implicit 'else' */
1096
1097         /* we only allow nat config for new conntracks */
1098         if (cda[CTA_NAT-1]) {
1099                 err = -EINVAL;
1100                 goto out_unlock;
1101         }
1102
1103         /* We manipulate the conntrack inside the global conntrack table lock,
1104          * so there's no need to increase the refcount */
1105         DEBUGP("conntrack found\n");
1106         err = -EEXIST;
1107         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1108                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1109
1110 out_unlock:
1111         write_unlock_bh(&ip_conntrack_lock);
1112         return err;
1113 }
1114
1115 /***********************************************************************
1116  * EXPECT
1117  ***********************************************************************/
1118
1119 static inline int
1120 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1121                          const struct ip_conntrack_tuple *tuple,
1122                          enum ctattr_expect type)
1123 {
1124         struct nfattr *nest_parms = NFA_NEST(skb, type);
1125
1126         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1127                 goto nfattr_failure;
1128
1129         NFA_NEST_END(skb, nest_parms);
1130
1131         return 0;
1132
1133 nfattr_failure:
1134         return -1;
1135 }
1136
1137 static inline int
1138 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1139                           const struct ip_conntrack_expect *exp)
1140 {
1141         struct ip_conntrack *master = exp->master;
1142         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1143         u_int32_t id = htonl(exp->id);
1144
1145         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1146                 goto nfattr_failure;
1147         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1148                 goto nfattr_failure;
1149         if (ctnetlink_exp_dump_tuple(skb,
1150                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1151                                  CTA_EXPECT_MASTER) < 0)
1152                 goto nfattr_failure;
1153
1154         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1155         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1156
1157         return 0;
1158
1159 nfattr_failure:
1160         return -1;
1161 }
1162
1163 static int
1164 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1165                     int event,
1166                     int nowait,
1167                     const struct ip_conntrack_expect *exp)
1168 {
1169         struct nlmsghdr *nlh;
1170         struct nfgenmsg *nfmsg;
1171         unsigned char *b;
1172
1173         b = skb->tail;
1174
1175         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1176         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1177         nfmsg  = NLMSG_DATA(nlh);
1178
1179         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1180         nfmsg->nfgen_family = AF_INET;
1181         nfmsg->version      = NFNETLINK_V0;
1182         nfmsg->res_id       = 0;
1183
1184         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1185                 goto nfattr_failure;
1186
1187         nlh->nlmsg_len = skb->tail - b;
1188         return skb->len;
1189
1190 nlmsg_failure:
1191 nfattr_failure:
1192         skb_trim(skb, b - skb->data);
1193         return -1;
1194 }
1195
1196 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1197 static int ctnetlink_expect_event(struct notifier_block *this,
1198                                   unsigned long events, void *ptr)
1199 {
1200         struct nlmsghdr *nlh;
1201         struct nfgenmsg *nfmsg;
1202         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1203         struct sk_buff *skb;
1204         unsigned int type;
1205         unsigned char *b;
1206         int flags = 0;
1207         u16 proto;
1208
1209         if (events & IPEXP_NEW) {
1210                 type = IPCTNL_MSG_EXP_NEW;
1211                 flags = NLM_F_CREATE|NLM_F_EXCL;
1212         } else
1213                 return NOTIFY_DONE;
1214
1215         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1216         if (!skb)
1217                 return NOTIFY_DONE;
1218
1219         b = skb->tail;
1220
1221         type |= NFNL_SUBSYS_CTNETLINK << 8;
1222         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1223         nfmsg = NLMSG_DATA(nlh);
1224
1225         nlh->nlmsg_flags    = flags;
1226         nfmsg->nfgen_family = AF_INET;
1227         nfmsg->version      = NFNETLINK_V0;
1228         nfmsg->res_id       = 0;
1229
1230         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1231                 goto nfattr_failure;
1232
1233         nlh->nlmsg_len = skb->tail - b;
1234         proto = exp->tuple.dst.protonum;
1235         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1236         return NOTIFY_DONE;
1237
1238 nlmsg_failure:
1239 nfattr_failure:
1240         kfree_skb(skb);
1241         return NOTIFY_DONE;
1242 }
1243 #endif
1244
1245 static int
1246 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1247 {
1248         struct ip_conntrack_expect *exp = NULL;
1249         struct list_head *i;
1250         u_int32_t *id = (u_int32_t *) &cb->args[0];
1251
1252         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1253
1254         read_lock_bh(&ip_conntrack_lock);
1255         list_for_each_prev(i, &ip_conntrack_expect_list) {
1256                 exp = (struct ip_conntrack_expect *) i;
1257                 if (exp->id <= *id)
1258                         continue;
1259                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1260                                             cb->nlh->nlmsg_seq,
1261                                             IPCTNL_MSG_EXP_NEW,
1262                                             1, exp) < 0)
1263                         goto out;
1264                 *id = exp->id;
1265         }
1266 out:
1267         read_unlock_bh(&ip_conntrack_lock);
1268
1269         DEBUGP("leaving, last id=%llu\n", *id);
1270
1271         return skb->len;
1272 }
1273
1274 static int
1275 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1276                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1277 {
1278         struct ip_conntrack_tuple tuple;
1279         struct ip_conntrack_expect *exp;
1280         struct sk_buff *skb2;
1281         int err = 0;
1282
1283         DEBUGP("entered %s\n", __FUNCTION__);
1284
1285         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1286                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1287                 u32 rlen;
1288
1289                 if (msg->nfgen_family != AF_INET)
1290                         return -EAFNOSUPPORT;
1291
1292                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1293                                                 ctnetlink_exp_dump_table,
1294                                                 ctnetlink_done)) != 0)
1295                         return -EINVAL;
1296                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1297                 if (rlen > skb->len)
1298                         rlen = skb->len;
1299                 skb_pull(skb, rlen);
1300                 return 0;
1301         }
1302
1303         if (cda[CTA_EXPECT_MASTER-1])
1304                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1305         else
1306                 return -EINVAL;
1307
1308         if (err < 0)
1309                 return err;
1310
1311         exp = ip_conntrack_expect_find(&tuple);
1312         if (!exp)
1313                 return -ENOENT;
1314
1315         err = -ENOMEM;
1316         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1317         if (!skb2)
1318                 goto out;
1319         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1320
1321         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1322                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1323                                       1, exp);
1324         if (err <= 0)
1325                 goto out;
1326
1327         ip_conntrack_expect_put(exp);
1328
1329         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1330         if (err < 0)
1331                 goto free;
1332
1333         return err;
1334
1335 out:
1336         ip_conntrack_expect_put(exp);
1337 free:
1338         if (skb2)
1339                 kfree_skb(skb2);
1340         return err;
1341 }
1342
1343 static int
1344 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1345                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1346 {
1347         struct ip_conntrack_expect *exp, *tmp;
1348         struct ip_conntrack_tuple tuple;
1349         struct ip_conntrack_helper *h;
1350         int err;
1351
1352         if (cda[CTA_EXPECT_TUPLE-1]) {
1353                 /* delete a single expect by tuple */
1354                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1355                 if (err < 0)
1356                         return err;
1357
1358                 /* bump usage count to 2 */
1359                 exp = ip_conntrack_expect_find(&tuple);
1360                 if (!exp)
1361                         return -ENOENT;
1362
1363                 if (cda[CTA_EXPECT_ID-1]) {
1364                         u_int32_t id =
1365                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1366                         if (exp->id != ntohl(id)) {
1367                                 ip_conntrack_expect_put(exp);
1368                                 return -ENOENT;
1369                         }
1370                 }
1371
1372                 /* after list removal, usage count == 1 */
1373                 ip_conntrack_unexpect_related(exp);
1374                 /* have to put what we 'get' above.
1375                  * after this line usage count == 0 */
1376                 ip_conntrack_expect_put(exp);
1377         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1378                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1379
1380                 /* delete all expectations for this helper */
1381                 write_lock_bh(&ip_conntrack_lock);
1382                 h = __ip_conntrack_helper_find_byname(name);
1383                 if (!h) {
1384                         write_unlock_bh(&ip_conntrack_lock);
1385                         return -EINVAL;
1386                 }
1387                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1388                                          list) {
1389                         if (exp->master->helper == h
1390                             && del_timer(&exp->timeout)) {
1391                                 ip_ct_unlink_expect(exp);
1392                                 ip_conntrack_expect_put(exp);
1393                         }
1394                 }
1395                 write_unlock(&ip_conntrack_lock);
1396         } else {
1397                 /* This basically means we have to flush everything*/
1398                 write_lock_bh(&ip_conntrack_lock);
1399                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1400                                          list) {
1401                         if (del_timer(&exp->timeout)) {
1402                                 ip_ct_unlink_expect(exp);
1403                                 ip_conntrack_expect_put(exp);
1404                         }
1405                 }
1406                 write_unlock_bh(&ip_conntrack_lock);
1407         }
1408
1409         return 0;
1410 }
1411 static int
1412 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1413 {
1414         return -EOPNOTSUPP;
1415 }
1416
1417 static int
1418 ctnetlink_create_expect(struct nfattr *cda[])
1419 {
1420         struct ip_conntrack_tuple tuple, mask, master_tuple;
1421         struct ip_conntrack_tuple_hash *h = NULL;
1422         struct ip_conntrack_expect *exp;
1423         struct ip_conntrack *ct;
1424         int err = 0;
1425
1426         DEBUGP("entered %s\n", __FUNCTION__);
1427
1428         /* caller guarantees that those three CTA_EXPECT_* exist */
1429         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1430         if (err < 0)
1431                 return err;
1432         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1433         if (err < 0)
1434                 return err;
1435         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1436         if (err < 0)
1437                 return err;
1438
1439         /* Look for master conntrack of this expectation */
1440         h = ip_conntrack_find_get(&master_tuple, NULL);
1441         if (!h)
1442                 return -ENOENT;
1443         ct = tuplehash_to_ctrack(h);
1444
1445         if (!ct->helper) {
1446                 /* such conntrack hasn't got any helper, abort */
1447                 err = -EINVAL;
1448                 goto out;
1449         }
1450
1451         exp = ip_conntrack_expect_alloc(ct);
1452         if (!exp) {
1453                 err = -ENOMEM;
1454                 goto out;
1455         }
1456
1457         exp->expectfn = NULL;
1458         exp->flags = 0;
1459         exp->master = ct;
1460         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1461         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1462
1463         err = ip_conntrack_expect_related(exp);
1464         ip_conntrack_expect_put(exp);
1465
1466 out:
1467         ip_conntrack_put(tuplehash_to_ctrack(h));
1468         return err;
1469 }
1470
1471 static int
1472 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1473                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1474 {
1475         struct ip_conntrack_tuple tuple;
1476         struct ip_conntrack_expect *exp;
1477         int err = 0;
1478
1479         DEBUGP("entered %s\n", __FUNCTION__);
1480
1481         if (!cda[CTA_EXPECT_TUPLE-1]
1482             || !cda[CTA_EXPECT_MASK-1]
1483             || !cda[CTA_EXPECT_MASTER-1])
1484                 return -EINVAL;
1485
1486         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1487         if (err < 0)
1488                 return err;
1489
1490         write_lock_bh(&ip_conntrack_lock);
1491         exp = __ip_conntrack_expect_find(&tuple);
1492
1493         if (!exp) {
1494                 write_unlock_bh(&ip_conntrack_lock);
1495                 err = -ENOENT;
1496                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1497                         err = ctnetlink_create_expect(cda);
1498                 return err;
1499         }
1500
1501         err = -EEXIST;
1502         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1503                 err = ctnetlink_change_expect(exp, cda);
1504         write_unlock_bh(&ip_conntrack_lock);
1505
1506         DEBUGP("leaving\n");
1507
1508         return err;
1509 }
1510
1511 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1512 static struct notifier_block ctnl_notifier = {
1513         .notifier_call  = ctnetlink_conntrack_event,
1514 };
1515
1516 static struct notifier_block ctnl_notifier_exp = {
1517         .notifier_call  = ctnetlink_expect_event,
1518 };
1519 #endif
1520
1521 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1522         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1523                                             .attr_count = CTA_MAX,
1524                                             .cap_required = CAP_NET_ADMIN },
1525         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1526                                             .attr_count = CTA_MAX,
1527                                             .cap_required = CAP_NET_ADMIN },
1528         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1529                                             .attr_count = CTA_MAX,
1530                                             .cap_required = CAP_NET_ADMIN },
1531         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1532                                             .attr_count = CTA_MAX,
1533                                             .cap_required = CAP_NET_ADMIN },
1534 };
1535
1536 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1537         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1538                                             .attr_count = CTA_EXPECT_MAX,
1539                                             .cap_required = CAP_NET_ADMIN },
1540         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1541                                             .attr_count = CTA_EXPECT_MAX,
1542                                             .cap_required = CAP_NET_ADMIN },
1543         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1544                                             .attr_count = CTA_EXPECT_MAX,
1545                                             .cap_required = CAP_NET_ADMIN },
1546 };
1547
1548 static struct nfnetlink_subsystem ctnl_subsys = {
1549         .name                           = "conntrack",
1550         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1551         .cb_count                       = IPCTNL_MSG_MAX,
1552         .cb                             = ctnl_cb,
1553 };
1554
1555 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1556         .name                           = "conntrack_expect",
1557         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1558         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1559         .cb                             = ctnl_exp_cb,
1560 };
1561
1562 static int __init ctnetlink_init(void)
1563 {
1564         int ret;
1565
1566         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1567         ret = nfnetlink_subsys_register(&ctnl_subsys);
1568         if (ret < 0) {
1569                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1570                 goto err_out;
1571         }
1572
1573         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1574         if (ret < 0) {
1575                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1576                 goto err_unreg_subsys;
1577         }
1578
1579 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1580         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1581         if (ret < 0) {
1582                 printk("ctnetlink_init: cannot register notifier.\n");
1583                 goto err_unreg_exp_subsys;
1584         }
1585
1586         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1587         if (ret < 0) {
1588                 printk("ctnetlink_init: cannot expect register notifier.\n");
1589                 goto err_unreg_notifier;
1590         }
1591 #endif
1592
1593         return 0;
1594
1595 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1596 err_unreg_notifier:
1597         ip_conntrack_unregister_notifier(&ctnl_notifier);
1598 err_unreg_exp_subsys:
1599         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1600 #endif
1601 err_unreg_subsys:
1602         nfnetlink_subsys_unregister(&ctnl_subsys);
1603 err_out:
1604         return ret;
1605 }
1606
1607 static void __exit ctnetlink_exit(void)
1608 {
1609         printk("ctnetlink: unregistering from nfnetlink.\n");
1610
1611 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1612         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1613         ip_conntrack_unregister_notifier(&ctnl_notifier);
1614 #endif
1615
1616         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1617         nfnetlink_subsys_unregister(&ctnl_subsys);
1618         return;
1619 }
1620
1621 module_init(ctnetlink_init);
1622 module_exit(ctnetlink_exit);