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
 486         if (nfattr_parse_nested(tb, CTA_IP_MAX, attr) < 0)
 487                 goto nfattr_failure;
 488
 489         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 490                 return -EINVAL;
 491
 492         if (!tb[CTA_IP_V4_SRC-1])
 493                 return -EINVAL;
 494         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 495
 496         if (!tb[CTA_IP_V4_DST-1])
 497                 return -EINVAL;
 498         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 499
 500         DEBUGP("leaving\n");
 501
 502         return 0;
 503
 504 nfattr_failure:
 505         return -1;
 506 }
 507
 508 static const int cta_min_proto[CTA_PROTO_MAX] = {
 509         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 510         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 511         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 512         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 513         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 514         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 515 };
 516
 517 static inline int
 518 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 519                             struct ip_conntrack_tuple *tuple)
 520 {
 521         struct nfattr *tb[CTA_PROTO_MAX];
 522         struct ip_conntrack_protocol *proto;
 523         int ret = 0;
 524
 525         DEBUGP("entered %s\n", __FUNCTION__);
 526
 527         if (nfattr_parse_nested(tb, CTA_PROTO_MAX, attr) < 0)
 528                 goto nfattr_failure;
 529
 530         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 531                 return -EINVAL;
 532
 533         if (!tb[CTA_PROTO_NUM-1])
 534                 return -EINVAL;
 535         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 536
 537         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 538
 539         if (likely(proto && proto->nfattr_to_tuple)) {
 540                 ret = proto->nfattr_to_tuple(tb, tuple);
 541                 ip_conntrack_proto_put(proto);
 542         }
 543
 544         return ret;
 545
 546 nfattr_failure:
 547         return -1;
 548 }
 549
 550 static inline int
 551 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 552                       enum ctattr_tuple type)
 553 {
 554         struct nfattr *tb[CTA_TUPLE_MAX];
 555         int err;
 556
 557         DEBUGP("entered %s\n", __FUNCTION__);
 558
 559         memset(tuple, 0, sizeof(*tuple));
 560
 561         if (nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]) < 0)
 562                 goto nfattr_failure;
 563
 564         if (!tb[CTA_TUPLE_IP-1])
 565                 return -EINVAL;
 566
 567         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 568         if (err < 0)
 569                 return err;
 570
 571         if (!tb[CTA_TUPLE_PROTO-1])
 572                 return -EINVAL;
 573
 574         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 575         if (err < 0)
 576                 return err;
 577
 578         /* orig and expect tuples get DIR_ORIGINAL */
 579         if (type == CTA_TUPLE_REPLY)
 580                 tuple->dst.dir = IP_CT_DIR_REPLY;
 581         else
 582                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 583
 584         DUMP_TUPLE(tuple);
 585
 586         DEBUGP("leaving\n");
 587
 588         return 0;
 589
 590 nfattr_failure:
 591         return -1;
 592 }
 593
 594 #ifdef CONFIG_IP_NF_NAT_NEEDED
 595 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
 596         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 597         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 598 };
 599
 600 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 601                                      const struct ip_conntrack *ct,
 602                                      struct ip_nat_range *range)
 603 {
 604         struct nfattr *tb[CTA_PROTONAT_MAX];
 605         struct ip_nat_protocol *npt;
 606
 607         DEBUGP("entered %s\n", __FUNCTION__);
 608
 609         if (nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr) < 0)
 610                 goto nfattr_failure;
 611
 612         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 613                 goto nfattr_failure;
 614
 615         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 616         if (!npt)
 617                 return 0;
 618
 619         if (!npt->nfattr_to_range) {
 620                 ip_nat_proto_put(npt);
 621                 return 0;
 622         }
 623
 624         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 625         if (npt->nfattr_to_range(tb, range) > 0)
 626                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 627
 628         ip_nat_proto_put(npt);
 629
 630         DEBUGP("leaving\n");
 631         return 0;
 632
 633 nfattr_failure:
 634         return -1;
 635 }
 636
 637 static inline int
 638 ctnetlink_parse_nat(struct nfattr *cda[],
 639                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 640 {
 641         struct nfattr *tb[CTA_NAT_MAX];
 642         int err;
 643
 644         DEBUGP("entered %s\n", __FUNCTION__);
 645
 646         memset(range, 0, sizeof(*range));
 647
 648         if (nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]) < 0)
 649                 goto nfattr_failure;
 650
 651         if (tb[CTA_NAT_MINIP-1])
 652                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 653
 654         if (!tb[CTA_NAT_MAXIP-1])
 655                 range->max_ip = range->min_ip;
 656         else
 657                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 658
 659         if (range->min_ip)
 660                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 661
 662         if (!tb[CTA_NAT_PROTO-1])
 663                 return 0;
 664
 665         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 666         if (err < 0)
 667                 return err;
 668
 669         DEBUGP("leaving\n");
 670         return 0;
 671
 672 nfattr_failure:
 673         return -1;
 674 }
 675 #endif
 676
 677 static inline int
 678 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 679 {
 680         struct nfattr *tb[CTA_HELP_MAX];
 681
 682         DEBUGP("entered %s\n", __FUNCTION__);
 683
 684         if (nfattr_parse_nested(tb, CTA_HELP_MAX, attr) < 0)
 685                 goto nfattr_failure;
 686
 687         if (!tb[CTA_HELP_NAME-1])
 688                 return -EINVAL;
 689
 690         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 691
 692         return 0;
 693
 694 nfattr_failure:
 695         return -1;
 696 }
 697
 698 static int
 699 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 700                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 701 {
 702         struct ip_conntrack_tuple_hash *h;
 703         struct ip_conntrack_tuple tuple;
 704         struct ip_conntrack *ct;
 705         int err = 0;
 706
 707         DEBUGP("entered %s\n", __FUNCTION__);
 708
 709         if (cda[CTA_TUPLE_ORIG-1])
 710                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 711         else if (cda[CTA_TUPLE_REPLY-1])
 712                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 713         else {
 714                 /* Flush the whole table */
 715                 ip_conntrack_flush();
 716                 return 0;
 717         }
 718
 719         if (err < 0)
 720                 return err;
 721
 722         h = ip_conntrack_find_get(&tuple, NULL);
 723         if (!h) {
 724                 DEBUGP("tuple not found in conntrack hash\n");
 725                 return -ENOENT;
 726         }
 727
 728         ct = tuplehash_to_ctrack(h);
 729
 730         if (cda[CTA_ID-1]) {
 731                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 732                 if (ct->id != id) {
 733                         ip_conntrack_put(ct);
 734                         return -ENOENT;
 735                 }
 736         }
 737         if (del_timer(&ct->timeout)) {
 738                 ip_conntrack_put(ct);
 739                 ct->timeout.function((unsigned long)ct);
 740                 return 0;
 741         }
 742         ip_conntrack_put(ct);
 743         DEBUGP("leaving\n");
 744
 745         return 0;
 746 }
 747
 748 static int
 749 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 750                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 751 {
 752         struct ip_conntrack_tuple_hash *h;
 753         struct ip_conntrack_tuple tuple;
 754         struct ip_conntrack *ct;
 755         struct sk_buff *skb2 = NULL;
 756         int err = 0;
 757
 758         DEBUGP("entered %s\n", __FUNCTION__);
 759
 760         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 761                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 762                 u32 rlen;
 763
 764                 if (msg->nfgen_family != AF_INET)
 765                         return -EAFNOSUPPORT;
 766
 767                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 768                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 769 #ifdef CONFIG_IP_NF_CT_ACCT
 770                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 771                                                 ctnetlink_dump_table_w,
 772                                                 ctnetlink_done)) != 0)
 773                                 return -EINVAL;
 774 #else
 775                         return -ENOTSUPP;
 776 #endif
 777                 } else {
 778                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 779                                                         ctnetlink_dump_table,
 780                                                         ctnetlink_done)) != 0)
 781                         return -EINVAL;
 782                 }
 783
 784                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 785                 if (rlen > skb->len)
 786                         rlen = skb->len;
 787                 skb_pull(skb, rlen);
 788                 return 0;
 789         }
 790
 791         if (cda[CTA_TUPLE_ORIG-1])
 792                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 793         else if (cda[CTA_TUPLE_REPLY-1])
 794                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 795         else
 796                 return -EINVAL;
 797
 798         if (err < 0)
 799                 return err;
 800
 801         h = ip_conntrack_find_get(&tuple, NULL);
 802         if (!h) {
 803                 DEBUGP("tuple not found in conntrack hash");
 804                 return -ENOENT;
 805         }
 806         DEBUGP("tuple found\n");
 807         ct = tuplehash_to_ctrack(h);
 808
 809         err = -ENOMEM;
 810         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 811         if (!skb2) {
 812                 ip_conntrack_put(ct);
 813                 return -ENOMEM;
 814         }
 815         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 816
 817         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 818                                   IPCTNL_MSG_CT_NEW, 1, ct);
 819         ip_conntrack_put(ct);
 820         if (err <= 0)
 821                 goto free;
 822
 823         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 824         if (err < 0)
 825                 goto out;
 826
 827         DEBUGP("leaving\n");
 828         return 0;
 829
 830 free:
 831         kfree_skb(skb2);
 832 out:
 833         return -1;
 834 }
 835
 836 static inline int
 837 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 838 {
 839         unsigned long d;
 840         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 841         d = ct->status ^ status;
 842
 843         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 844                 /* unchangeable */
 845                 return -EINVAL;
 846
 847         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 848                 /* SEEN_REPLY bit can only be set */
 849                 return -EINVAL;
 850
 851
 852         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 853                 /* ASSURED bit can only be set */
 854                 return -EINVAL;
 855
 856         if (cda[CTA_NAT-1]) {
 857 #ifndef CONFIG_IP_NF_NAT_NEEDED
 858                 return -EINVAL;
 859 #else
 860                 unsigned int hooknum;
 861                 struct ip_nat_range range;
 862
 863                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 864                         return -EINVAL;
 865
 866                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 867                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 868                        htons(range.min.all), htons(range.max.all));
 869
 870                 /* This is tricky but it works. ip_nat_setup_info needs the
 871                  * hook number as parameter, so let's do the correct
 872                  * conversion and run away */
 873                 if (status & IPS_SRC_NAT_DONE)
 874                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 875                 else if (status & IPS_DST_NAT_DONE)
 876                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 877                 else
 878                         return -EINVAL; /* Missing NAT flags */
 879
 880                 DEBUGP("NAT status: %lu\n",
 881                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 882
 883                 if (ip_nat_initialized(ct, hooknum))
 884                         return -EEXIST;
 885                 ip_nat_setup_info(ct, &range, hooknum);
 886
 887                 DEBUGP("NAT status after setup_info: %lu\n",
 888                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 889 #endif
 890         }
 891
 892         /* Be careful here, modifying NAT bits can screw up things,
 893          * so don't let users modify them directly if they don't pass
 894          * ip_nat_range. */
 895         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 896         return 0;
 897 }
 898
 899
 900 static inline int
 901 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 902 {
 903         struct ip_conntrack_helper *helper;
 904         char *helpname;
 905         int err;
 906
 907         DEBUGP("entered %s\n", __FUNCTION__);
 908
 909         /* don't change helper of sibling connections */
 910         if (ct->master)
 911                 return -EINVAL;
 912
 913         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 914         if (err < 0)
 915                 return err;
 916
 917         helper = __ip_conntrack_helper_find_byname(helpname);
 918         if (!helper) {
 919                 if (!strcmp(helpname, ""))
 920                         helper = NULL;
 921                 else
 922                         return -EINVAL;
 923         }
 924
 925         if (ct->helper) {
 926                 if (!helper) {
 927                         /* we had a helper before ... */
 928                         ip_ct_remove_expectations(ct);
 929                         ct->helper = NULL;
 930                 } else {
 931                         /* need to zero data of old helper */
 932                         memset(&ct->help, 0, sizeof(ct->help));
 933                 }
 934         }
 935
 936         ct->helper = helper;
 937
 938         return 0;
 939 }
 940
 941 static inline int
 942 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 943 {
 944         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 945
 946         if (!del_timer(&ct->timeout))
 947                 return -ETIME;
 948
 949         ct->timeout.expires = jiffies + timeout * HZ;
 950         add_timer(&ct->timeout);
 951
 952         return 0;
 953 }
 954
 955 static inline int
 956 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 957 {
 958         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 959         struct ip_conntrack_protocol *proto;
 960         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 961         int err = 0;
 962
 963         if (nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr) < 0)
 964                 goto nfattr_failure;
 965
 966         proto = ip_conntrack_proto_find_get(npt);
 967         if (!proto)
 968                 return -EINVAL;
 969
 970         if (proto->from_nfattr)
 971                 err = proto->from_nfattr(tb, ct);
 972         ip_conntrack_proto_put(proto);
 973
 974         return err;
 975
 976 nfattr_failure:
 977         return -ENOMEM;
 978 }
 979
 980 static int
 981 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 982 {
 983         int err;
 984
 985         DEBUGP("entered %s\n", __FUNCTION__);
 986
 987         if (cda[CTA_HELP-1]) {
 988                 err = ctnetlink_change_helper(ct, cda);
 989                 if (err < 0)
 990                         return err;
 991         }
 992
 993         if (cda[CTA_TIMEOUT-1]) {
 994                 err = ctnetlink_change_timeout(ct, cda);
 995                 if (err < 0)
 996                         return err;
 997         }
 998
 999         if (cda[CTA_STATUS-1]) {
1000                 err = ctnetlink_change_status(ct, cda);
1001                 if (err < 0)
1002                         return err;
1003         }
1004
1005         if (cda[CTA_PROTOINFO-1]) {
1006                 err = ctnetlink_change_protoinfo(ct, cda);
1007                 if (err < 0)
1008                         return err;
1009         }
1010
1011         DEBUGP("all done\n");
1012         return 0;
1013 }
1014
1015 static int
1016 ctnetlink_create_conntrack(struct nfattr *cda[],
1017                            struct ip_conntrack_tuple *otuple,
1018                            struct ip_conntrack_tuple *rtuple)
1019 {
1020         struct ip_conntrack *ct;
1021         int err = -EINVAL;
1022
1023         DEBUGP("entered %s\n", __FUNCTION__);
1024
1025         ct = ip_conntrack_alloc(otuple, rtuple);
1026         if (ct == NULL || IS_ERR(ct))
1027                 return -ENOMEM;
1028
1029         if (!cda[CTA_TIMEOUT-1])
1030                 goto err;
1031         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1032
1033         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1034         ct->status |= IPS_CONFIRMED;
1035
1036         err = ctnetlink_change_status(ct, cda);
1037         if (err < 0)
1038                 goto err;
1039
1040         if (cda[CTA_PROTOINFO-1]) {
1041                 err = ctnetlink_change_protoinfo(ct, cda);
1042                 if (err < 0)
1043                         return err;
1044         }
1045
1046         ct->helper = ip_conntrack_helper_find_get(rtuple);
1047
1048         add_timer(&ct->timeout);
1049         ip_conntrack_hash_insert(ct);
1050
1051         if (ct->helper)
1052                 ip_conntrack_helper_put(ct->helper);
1053
1054         DEBUGP("conntrack with id %u inserted\n", ct->id);
1055         return 0;
1056
1057 err:
1058         ip_conntrack_free(ct);
1059         return err;
1060 }
1061
1062 static int
1063 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1064                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1065 {
1066         struct ip_conntrack_tuple otuple, rtuple;
1067         struct ip_conntrack_tuple_hash *h = NULL;
1068         int err = 0;
1069
1070         DEBUGP("entered %s\n", __FUNCTION__);
1071
1072         if (cda[CTA_TUPLE_ORIG-1]) {
1073                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1074                 if (err < 0)
1075                         return err;
1076         }
1077
1078         if (cda[CTA_TUPLE_REPLY-1]) {
1079                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1080                 if (err < 0)
1081                         return err;
1082         }
1083
1084         write_lock_bh(&ip_conntrack_lock);
1085         if (cda[CTA_TUPLE_ORIG-1])
1086                 h = __ip_conntrack_find(&otuple, NULL);
1087         else if (cda[CTA_TUPLE_REPLY-1])
1088                 h = __ip_conntrack_find(&rtuple, NULL);
1089
1090         if (h == NULL) {
1091                 write_unlock_bh(&ip_conntrack_lock);
1092                 DEBUGP("no such conntrack, create new\n");
1093                 err = -ENOENT;
1094                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1095                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1096                 return err;
1097         }
1098         /* implicit 'else' */
1099
1100         /* we only allow nat config for new conntracks */
1101         if (cda[CTA_NAT-1]) {
1102                 err = -EINVAL;
1103                 goto out_unlock;
1104         }
1105
1106         /* We manipulate the conntrack inside the global conntrack table lock,
1107          * so there's no need to increase the refcount */
1108         DEBUGP("conntrack found\n");
1109         err = -EEXIST;
1110         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1111                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1112
1113 out_unlock:
1114         write_unlock_bh(&ip_conntrack_lock);
1115         return err;
1116 }
1117
1118 /***********************************************************************
1119  * EXPECT
1120  ***********************************************************************/
1121
1122 static inline int
1123 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1124                          const struct ip_conntrack_tuple *tuple,
1125                          enum ctattr_expect type)
1126 {
1127         struct nfattr *nest_parms = NFA_NEST(skb, type);
1128
1129         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1130                 goto nfattr_failure;
1131
1132         NFA_NEST_END(skb, nest_parms);
1133
1134         return 0;
1135
1136 nfattr_failure:
1137         return -1;
1138 }
1139
1140 static inline int
1141 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1142                           const struct ip_conntrack_expect *exp)
1143 {
1144         struct ip_conntrack *master = exp->master;
1145         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1146         u_int32_t id = htonl(exp->id);
1147
1148         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1149                 goto nfattr_failure;
1150         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1151                 goto nfattr_failure;
1152         if (ctnetlink_exp_dump_tuple(skb,
1153                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1154                                  CTA_EXPECT_MASTER) < 0)
1155                 goto nfattr_failure;
1156
1157         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1158         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1159
1160         return 0;
1161
1162 nfattr_failure:
1163         return -1;
1164 }
1165
1166 static int
1167 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1168                     int event,
1169                     int nowait,
1170                     const struct ip_conntrack_expect *exp)
1171 {
1172         struct nlmsghdr *nlh;
1173         struct nfgenmsg *nfmsg;
1174         unsigned char *b;
1175
1176         b = skb->tail;
1177
1178         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1179         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1180         nfmsg  = NLMSG_DATA(nlh);
1181
1182         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1183         nfmsg->nfgen_family = AF_INET;
1184         nfmsg->version      = NFNETLINK_V0;
1185         nfmsg->res_id       = 0;
1186
1187         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1188                 goto nfattr_failure;
1189
1190         nlh->nlmsg_len = skb->tail - b;
1191         return skb->len;
1192
1193 nlmsg_failure:
1194 nfattr_failure:
1195         skb_trim(skb, b - skb->data);
1196         return -1;
1197 }
1198
1199 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1200 static int ctnetlink_expect_event(struct notifier_block *this,
1201                                   unsigned long events, void *ptr)
1202 {
1203         struct nlmsghdr *nlh;
1204         struct nfgenmsg *nfmsg;
1205         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1206         struct sk_buff *skb;
1207         unsigned int type;
1208         unsigned char *b;
1209         int flags = 0;
1210         u16 proto;
1211
1212         if (events & IPEXP_NEW) {
1213                 type = IPCTNL_MSG_EXP_NEW;
1214                 flags = NLM_F_CREATE|NLM_F_EXCL;
1215         } else
1216                 return NOTIFY_DONE;
1217
1218         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1219         if (!skb)
1220                 return NOTIFY_DONE;
1221
1222         b = skb->tail;
1223
1224         type |= NFNL_SUBSYS_CTNETLINK << 8;
1225         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1226         nfmsg = NLMSG_DATA(nlh);
1227
1228         nlh->nlmsg_flags    = flags;
1229         nfmsg->nfgen_family = AF_INET;
1230         nfmsg->version      = NFNETLINK_V0;
1231         nfmsg->res_id       = 0;
1232
1233         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1234                 goto nfattr_failure;
1235
1236         nlh->nlmsg_len = skb->tail - b;
1237         proto = exp->tuple.dst.protonum;
1238         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1239         return NOTIFY_DONE;
1240
1241 nlmsg_failure:
1242 nfattr_failure:
1243         kfree_skb(skb);
1244         return NOTIFY_DONE;
1245 }
1246 #endif
1247
1248 static int
1249 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1250 {
1251         struct ip_conntrack_expect *exp = NULL;
1252         struct list_head *i;
1253         u_int32_t *id = (u_int32_t *) &cb->args[0];
1254
1255         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1256
1257         read_lock_bh(&ip_conntrack_lock);
1258         list_for_each_prev(i, &ip_conntrack_expect_list) {
1259                 exp = (struct ip_conntrack_expect *) i;
1260                 if (exp->id <= *id)
1261                         continue;
1262                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1263                                             cb->nlh->nlmsg_seq,
1264                                             IPCTNL_MSG_EXP_NEW,
1265                                             1, exp) < 0)
1266                         goto out;
1267                 *id = exp->id;
1268         }
1269 out:
1270         read_unlock_bh(&ip_conntrack_lock);
1271
1272         DEBUGP("leaving, last id=%llu\n", *id);
1273
1274         return skb->len;
1275 }
1276
1277 static int
1278 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1279                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1280 {
1281         struct ip_conntrack_tuple tuple;
1282         struct ip_conntrack_expect *exp;
1283         struct sk_buff *skb2;
1284         int err = 0;
1285
1286         DEBUGP("entered %s\n", __FUNCTION__);
1287
1288         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1289                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1290                 u32 rlen;
1291
1292                 if (msg->nfgen_family != AF_INET)
1293                         return -EAFNOSUPPORT;
1294
1295                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1296                                                 ctnetlink_exp_dump_table,
1297                                                 ctnetlink_done)) != 0)
1298                         return -EINVAL;
1299                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1300                 if (rlen > skb->len)
1301                         rlen = skb->len;
1302                 skb_pull(skb, rlen);
1303                 return 0;
1304         }
1305
1306         if (cda[CTA_EXPECT_MASTER-1])
1307                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1308         else
1309                 return -EINVAL;
1310
1311         if (err < 0)
1312                 return err;
1313
1314         exp = ip_conntrack_expect_find(&tuple);
1315         if (!exp)
1316                 return -ENOENT;
1317
1318         err = -ENOMEM;
1319         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1320         if (!skb2)
1321                 goto out;
1322         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1323
1324         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1325                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1326                                       1, exp);
1327         if (err <= 0)
1328                 goto free;
1329
1330         ip_conntrack_expect_put(exp);
1331
1332         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1333
1334 free:
1335         kfree_skb(skb2);
1336 out:
1337         ip_conntrack_expect_put(exp);
1338         return err;
1339 }
1340
1341 static int
1342 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1343                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1344 {
1345         struct ip_conntrack_expect *exp, *tmp;
1346         struct ip_conntrack_tuple tuple;
1347         struct ip_conntrack_helper *h;
1348         int err;
1349
1350         if (cda[CTA_EXPECT_TUPLE-1]) {
1351                 /* delete a single expect by tuple */
1352                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1353                 if (err < 0)
1354                         return err;
1355
1356                 /* bump usage count to 2 */
1357                 exp = ip_conntrack_expect_find(&tuple);
1358                 if (!exp)
1359                         return -ENOENT;
1360
1361                 if (cda[CTA_EXPECT_ID-1]) {
1362                         u_int32_t id =
1363                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1364                         if (exp->id != ntohl(id)) {
1365                                 ip_conntrack_expect_put(exp);
1366                                 return -ENOENT;
1367                         }
1368                 }
1369
1370                 /* after list removal, usage count == 1 */
1371                 ip_conntrack_unexpect_related(exp);
1372                 /* have to put what we 'get' above.
1373                  * after this line usage count == 0 */
1374                 ip_conntrack_expect_put(exp);
1375         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1376                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1377
1378                 /* delete all expectations for this helper */
1379                 write_lock_bh(&ip_conntrack_lock);
1380                 h = __ip_conntrack_helper_find_byname(name);
1381                 if (!h) {
1382                         write_unlock_bh(&ip_conntrack_lock);
1383                         return -EINVAL;
1384                 }
1385                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1386                                          list) {
1387                         if (exp->master->helper == h
1388                             && del_timer(&exp->timeout)) {
1389                                 ip_ct_unlink_expect(exp);
1390                                 ip_conntrack_expect_put(exp);
1391                         }
1392                 }
1393                 write_unlock(&ip_conntrack_lock);
1394         } else {
1395                 /* This basically means we have to flush everything*/
1396                 write_lock_bh(&ip_conntrack_lock);
1397                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1398                                          list) {
1399                         if (del_timer(&exp->timeout)) {
1400                                 ip_ct_unlink_expect(exp);
1401                                 ip_conntrack_expect_put(exp);
1402                         }
1403                 }
1404                 write_unlock_bh(&ip_conntrack_lock);
1405         }
1406
1407         return 0;
1408 }
1409 static int
1410 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1411 {
1412         return -EOPNOTSUPP;
1413 }
1414
1415 static int
1416 ctnetlink_create_expect(struct nfattr *cda[])
1417 {
1418         struct ip_conntrack_tuple tuple, mask, master_tuple;
1419         struct ip_conntrack_tuple_hash *h = NULL;
1420         struct ip_conntrack_expect *exp;
1421         struct ip_conntrack *ct;
1422         int err = 0;
1423
1424         DEBUGP("entered %s\n", __FUNCTION__);
1425
1426         /* caller guarantees that those three CTA_EXPECT_* exist */
1427         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1428         if (err < 0)
1429                 return err;
1430         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1431         if (err < 0)
1432                 return err;
1433         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1434         if (err < 0)
1435                 return err;
1436
1437         /* Look for master conntrack of this expectation */
1438         h = ip_conntrack_find_get(&master_tuple, NULL);
1439         if (!h)
1440                 return -ENOENT;
1441         ct = tuplehash_to_ctrack(h);
1442
1443         if (!ct->helper) {
1444                 /* such conntrack hasn't got any helper, abort */
1445                 err = -EINVAL;
1446                 goto out;
1447         }
1448
1449         exp = ip_conntrack_expect_alloc(ct);
1450         if (!exp) {
1451                 err = -ENOMEM;
1452                 goto out;
1453         }
1454
1455         exp->expectfn = NULL;
1456         exp->flags = 0;
1457         exp->master = ct;
1458         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1459         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1460
1461         err = ip_conntrack_expect_related(exp);
1462         ip_conntrack_expect_put(exp);
1463
1464 out:
1465         ip_conntrack_put(tuplehash_to_ctrack(h));
1466         return err;
1467 }
1468
1469 static int
1470 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1471                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1472 {
1473         struct ip_conntrack_tuple tuple;
1474         struct ip_conntrack_expect *exp;
1475         int err = 0;
1476
1477         DEBUGP("entered %s\n", __FUNCTION__);
1478
1479         if (!cda[CTA_EXPECT_TUPLE-1]
1480             || !cda[CTA_EXPECT_MASK-1]
1481             || !cda[CTA_EXPECT_MASTER-1])
1482                 return -EINVAL;
1483
1484         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1485         if (err < 0)
1486                 return err;
1487
1488         write_lock_bh(&ip_conntrack_lock);
1489         exp = __ip_conntrack_expect_find(&tuple);
1490
1491         if (!exp) {
1492                 write_unlock_bh(&ip_conntrack_lock);
1493                 err = -ENOENT;
1494                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1495                         err = ctnetlink_create_expect(cda);
1496                 return err;
1497         }
1498
1499         err = -EEXIST;
1500         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1501                 err = ctnetlink_change_expect(exp, cda);
1502         write_unlock_bh(&ip_conntrack_lock);
1503
1504         DEBUGP("leaving\n");
1505
1506         return err;
1507 }
1508
1509 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1510 static struct notifier_block ctnl_notifier = {
1511         .notifier_call  = ctnetlink_conntrack_event,
1512 };
1513
1514 static struct notifier_block ctnl_notifier_exp = {
1515         .notifier_call  = ctnetlink_expect_event,
1516 };
1517 #endif
1518
1519 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1520         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1521                                             .attr_count = CTA_MAX,
1522                                             .cap_required = CAP_NET_ADMIN },
1523         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1524                                             .attr_count = CTA_MAX,
1525                                             .cap_required = CAP_NET_ADMIN },
1526         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1527                                             .attr_count = CTA_MAX,
1528                                             .cap_required = CAP_NET_ADMIN },
1529         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1530                                             .attr_count = CTA_MAX,
1531                                             .cap_required = CAP_NET_ADMIN },
1532 };
1533
1534 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1535         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1536                                             .attr_count = CTA_EXPECT_MAX,
1537                                             .cap_required = CAP_NET_ADMIN },
1538         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1539                                             .attr_count = CTA_EXPECT_MAX,
1540                                             .cap_required = CAP_NET_ADMIN },
1541         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1542                                             .attr_count = CTA_EXPECT_MAX,
1543                                             .cap_required = CAP_NET_ADMIN },
1544 };
1545
1546 static struct nfnetlink_subsystem ctnl_subsys = {
1547         .name                           = "conntrack",
1548         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1549         .cb_count                       = IPCTNL_MSG_MAX,
1550         .cb                             = ctnl_cb,
1551 };
1552
1553 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1554         .name                           = "conntrack_expect",
1555         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1556         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1557         .cb                             = ctnl_exp_cb,
1558 };
1559
1560 static int __init ctnetlink_init(void)
1561 {
1562         int ret;
1563
1564         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1565         ret = nfnetlink_subsys_register(&ctnl_subsys);
1566         if (ret < 0) {
1567                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1568                 goto err_out;
1569         }
1570
1571         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1572         if (ret < 0) {
1573                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1574                 goto err_unreg_subsys;
1575         }
1576
1577 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1578         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1579         if (ret < 0) {
1580                 printk("ctnetlink_init: cannot register notifier.\n");
1581                 goto err_unreg_exp_subsys;
1582         }
1583
1584         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1585         if (ret < 0) {
1586                 printk("ctnetlink_init: cannot expect register notifier.\n");
1587                 goto err_unreg_notifier;
1588         }
1589 #endif
1590
1591         return 0;
1592
1593 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1594 err_unreg_notifier:
1595         ip_conntrack_unregister_notifier(&ctnl_notifier);
1596 err_unreg_exp_subsys:
1597         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1598 #endif
1599 err_unreg_subsys:
1600         nfnetlink_subsys_unregister(&ctnl_subsys);
1601 err_out:
1602         return ret;
1603 }
1604
1605 static void __exit ctnetlink_exit(void)
1606 {
1607         printk("ctnetlink: unregistering from nfnetlink.\n");
1608
1609 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1610         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1611         ip_conntrack_unregister_notifier(&ctnl_notifier);
1612 #endif
1613
1614         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1615         nfnetlink_subsys_unregister(&ctnl_subsys);
1616         return;
1617 }
1618
1619 module_init(ctnetlink_init);
1620 module_exit(ctnetlink_exit);