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
  32 #include <linux/netfilter.h>
  33 #include <linux/netfilter_ipv4/ip_conntrack.h>
  34 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  37 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  38
  39 #include <linux/netfilter/nfnetlink.h>
  40 #include <linux/netfilter/nfnetlink_conntrack.h>
  41
  42 MODULE_LICENSE("GPL");
  43
  44 static char __initdata version[] = "0.90";
  45
  46 #if 0
  47 #define DEBUGP printk
  48 #else
  49 #define DEBUGP(format, args...)
  50 #endif
  51
  52
  53 static inline int
  54 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  55                             const struct ip_conntrack_tuple *tuple)
  56 {
  57         struct ip_conntrack_protocol *proto;
  58         int ret = 0;
  59
  60         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  61
  62         /* If no protocol helper is found, this function will return the
  63          * generic protocol helper, so proto won't *ever* be NULL */
  64         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  65         if (likely(proto->tuple_to_nfattr))
  66                 ret = proto->tuple_to_nfattr(skb, tuple);
  67
  68         ip_conntrack_proto_put(proto);
  69
  70         return ret;
  71
  72 nfattr_failure:
  73         return -1;
  74 }
  75
  76 static inline int
  77 ctnetlink_dump_tuples(struct sk_buff *skb,
  78                       const struct ip_conntrack_tuple *tuple)
  79 {
  80         struct nfattr *nest_parms;
  81
  82         nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  83         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  84         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  85         NFA_NEST_END(skb, nest_parms);
  86
  87         nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  88         ctnetlink_dump_tuples_proto(skb, tuple);
  89         NFA_NEST_END(skb, nest_parms);
  90
  91         return 0;
  92
  93 nfattr_failure:
  94         return -1;
  95 }
  96
  97 static inline int
  98 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
  99 {
 100         u_int32_t status = htonl((u_int32_t) ct->status);
 101         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 102         return 0;
 103
 104 nfattr_failure:
 105         return -1;
 106 }
 107
 108 static inline int
 109 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 110 {
 111         long timeout_l = ct->timeout.expires - jiffies;
 112         u_int32_t timeout;
 113
 114         if (timeout_l < 0)
 115                 timeout = 0;
 116         else
 117                 timeout = htonl(timeout_l / HZ);
 118
 119         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 120         return 0;
 121
 122 nfattr_failure:
 123         return -1;
 124 }
 125
 126 static inline int
 127 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 128 {
 129         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 130
 131         struct nfattr *nest_proto;
 132         int ret;
 133
 134         if (!proto->to_nfattr) {
 135                 ip_conntrack_proto_put(proto);
 136                 return 0;
 137         }
 138
 139         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 140
 141         ret = proto->to_nfattr(skb, nest_proto, ct);
 142
 143         ip_conntrack_proto_put(proto);
 144
 145         NFA_NEST_END(skb, nest_proto);
 146
 147         return ret;
 148
 149 nfattr_failure:
 150         return -1;
 151 }
 152
 153 static inline int
 154 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 155 {
 156         struct nfattr *nest_helper;
 157
 158         if (!ct->helper)
 159                 return 0;
 160
 161         nest_helper = NFA_NEST(skb, CTA_HELP);
 162         NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
 163
 164         if (ct->helper->to_nfattr)
 165                 ct->helper->to_nfattr(skb, ct);
 166
 167         NFA_NEST_END(skb, nest_helper);
 168
 169         return 0;
 170
 171 nfattr_failure:
 172         return -1;
 173 }
 174
 175 #ifdef CONFIG_IP_NF_CT_ACCT
 176 static inline int
 177 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 178                         enum ip_conntrack_dir dir)
 179 {
 180         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 181         struct nfattr *nest_count = NFA_NEST(skb, type);
 182         u_int32_t tmp;
 183
 184         tmp = htonl(ct->counters[dir].packets);
 185         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 186
 187         tmp = htonl(ct->counters[dir].bytes);
 188         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 189
 190         NFA_NEST_END(skb, nest_count);
 191
 192         return 0;
 193
 194 nfattr_failure:
 195         return -1;
 196 }
 197 #else
 198 #define ctnetlink_dump_counters(a, b, c) (0)
 199 #endif
 200
 201 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 202 static inline int
 203 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 204 {
 205         u_int32_t mark = htonl(ct->mark);
 206
 207         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 208         return 0;
 209
 210 nfattr_failure:
 211         return -1;
 212 }
 213 #else
 214 #define ctnetlink_dump_mark(a, b) (0)
 215 #endif
 216
 217 static inline int
 218 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 219 {
 220         u_int32_t id = htonl(ct->id);
 221         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 222         return 0;
 223
 224 nfattr_failure:
 225         return -1;
 226 }
 227
 228 static inline int
 229 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 230 {
 231         unsigned int use = htonl(atomic_read(&ct->ct_general.use));
 232
 233         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 234         return 0;
 235
 236 nfattr_failure:
 237         return -1;
 238 }
 239
 240 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 241
 242 static int
 243 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 244                     int event, int nowait,
 245                     const struct ip_conntrack *ct)
 246 {
 247         struct nlmsghdr *nlh;
 248         struct nfgenmsg *nfmsg;
 249         struct nfattr *nest_parms;
 250         unsigned char *b;
 251
 252         b = skb->tail;
 253
 254         event |= NFNL_SUBSYS_CTNETLINK << 8;
 255         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 256         nfmsg  = NLMSG_DATA(nlh);
 257
 258         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 259         nfmsg->nfgen_family = AF_INET;
 260         nfmsg->version      = NFNETLINK_V0;
 261         nfmsg->res_id       = 0;
 262
 263         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 264         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 265                 goto nfattr_failure;
 266         NFA_NEST_END(skb, nest_parms);
 267
 268         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 269         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 270                 goto nfattr_failure;
 271         NFA_NEST_END(skb, nest_parms);
 272
 273         if (ctnetlink_dump_status(skb, ct) < 0 ||
 274             ctnetlink_dump_timeout(skb, ct) < 0 ||
 275             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 276             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 277             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 278             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 279             ctnetlink_dump_mark(skb, ct) < 0 ||
 280             ctnetlink_dump_id(skb, ct) < 0 ||
 281             ctnetlink_dump_use(skb, ct) < 0)
 282                 goto nfattr_failure;
 283
 284         nlh->nlmsg_len = skb->tail - b;
 285         return skb->len;
 286
 287 nlmsg_failure:
 288 nfattr_failure:
 289         skb_trim(skb, b - skb->data);
 290         return -1;
 291 }
 292
 293 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 294 static int ctnetlink_conntrack_event(struct notifier_block *this,
 295                                      unsigned long events, void *ptr)
 296 {
 297         struct nlmsghdr *nlh;
 298         struct nfgenmsg *nfmsg;
 299         struct nfattr *nest_parms;
 300         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 301         struct sk_buff *skb;
 302         unsigned int type;
 303         unsigned char *b;
 304         unsigned int flags = 0, group;
 305
 306         /* ignore our fake conntrack entry */
 307         if (ct == &ip_conntrack_untracked)
 308                 return NOTIFY_DONE;
 309
 310         if (events & IPCT_DESTROY) {
 311                 type = IPCTNL_MSG_CT_DELETE;
 312                 group = NFNLGRP_CONNTRACK_DESTROY;
 313                 goto alloc_skb;
 314         }
 315         if (events & (IPCT_NEW | IPCT_RELATED)) {
 316                 type = IPCTNL_MSG_CT_NEW;
 317                 flags = NLM_F_CREATE|NLM_F_EXCL;
 318                 /* dump everything */
 319                 events = ~0UL;
 320                 group = NFNLGRP_CONNTRACK_NEW;
 321                 goto alloc_skb;
 322         }
 323         if (events & (IPCT_STATUS |
 324                       IPCT_PROTOINFO |
 325                       IPCT_HELPER |
 326                       IPCT_HELPINFO |
 327                       IPCT_NATINFO)) {
 328                 type = IPCTNL_MSG_CT_NEW;
 329                 group = NFNLGRP_CONNTRACK_UPDATE;
 330                 goto alloc_skb;
 331         }
 332
 333         return NOTIFY_DONE;
 334
 335 alloc_skb:
 336   /* FIXME: Check if there are any listeners before, don't hurt performance */
 337
 338         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 339         if (!skb)
 340                 return NOTIFY_DONE;
 341
 342         b = skb->tail;
 343
 344         type |= NFNL_SUBSYS_CTNETLINK << 8;
 345         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 346         nfmsg = NLMSG_DATA(nlh);
 347
 348         nlh->nlmsg_flags    = flags;
 349         nfmsg->nfgen_family = AF_INET;
 350         nfmsg->version  = NFNETLINK_V0;
 351         nfmsg->res_id   = 0;
 352
 353         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 354         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 355                 goto nfattr_failure;
 356         NFA_NEST_END(skb, nest_parms);
 357
 358         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 359         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 360                 goto nfattr_failure;
 361         NFA_NEST_END(skb, nest_parms);
 362
 363         /* NAT stuff is now a status flag */
 364         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 365             && ctnetlink_dump_status(skb, ct) < 0)
 366                 goto nfattr_failure;
 367         if (events & IPCT_REFRESH
 368             && ctnetlink_dump_timeout(skb, ct) < 0)
 369                 goto nfattr_failure;
 370         if (events & IPCT_PROTOINFO
 371             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 372                 goto nfattr_failure;
 373         if (events & IPCT_HELPINFO
 374             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 375                 goto nfattr_failure;
 376
 377         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 378             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 379                 goto nfattr_failure;
 380
 381         nlh->nlmsg_len = skb->tail - b;
 382         nfnetlink_send(skb, 0, group, 0);
 383         return NOTIFY_DONE;
 384
 385 nlmsg_failure:
 386 nfattr_failure:
 387         kfree_skb(skb);
 388         return NOTIFY_DONE;
 389 }
 390 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 391
 392 static int ctnetlink_done(struct netlink_callback *cb)
 393 {
 394         DEBUGP("entered %s\n", __FUNCTION__);
 395         return 0;
 396 }
 397
 398 static int
 399 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 400 {
 401         struct ip_conntrack *ct = NULL;
 402         struct ip_conntrack_tuple_hash *h;
 403         struct list_head *i;
 404         u_int32_t *id = (u_int32_t *) &cb->args[1];
 405
 406         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 407                         cb->args[0], *id);
 408
 409         read_lock_bh(&ip_conntrack_lock);
 410         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 411                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 412                         h = (struct ip_conntrack_tuple_hash *) i;
 413                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 414                                 continue;
 415                         ct = tuplehash_to_ctrack(h);
 416                         if (ct->id <= *id)
 417                                 continue;
 418                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 419                                                 cb->nlh->nlmsg_seq,
 420                                                 IPCTNL_MSG_CT_NEW,
 421                                                 1, ct) < 0)
 422                                 goto out;
 423                         *id = ct->id;
 424                 }
 425         }
 426 out:
 427         read_unlock_bh(&ip_conntrack_lock);
 428
 429         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 430
 431         return skb->len;
 432 }
 433
 434 #ifdef CONFIG_IP_NF_CT_ACCT
 435 static int
 436 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 437 {
 438         struct ip_conntrack *ct = NULL;
 439         struct ip_conntrack_tuple_hash *h;
 440         struct list_head *i;
 441         u_int32_t *id = (u_int32_t *) &cb->args[1];
 442
 443         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 444                         cb->args[0], *id);
 445
 446         write_lock_bh(&ip_conntrack_lock);
 447         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 448                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 449                         h = (struct ip_conntrack_tuple_hash *) i;
 450                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 451                                 continue;
 452                         ct = tuplehash_to_ctrack(h);
 453                         if (ct->id <= *id)
 454                                 continue;
 455                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 456                                                 cb->nlh->nlmsg_seq,
 457                                                 IPCTNL_MSG_CT_NEW,
 458                                                 1, ct) < 0)
 459                                 goto out;
 460                         *id = ct->id;
 461
 462                         memset(&ct->counters, 0, sizeof(ct->counters));
 463                 }
 464         }
 465 out:
 466         write_unlock_bh(&ip_conntrack_lock);
 467
 468         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 469
 470         return skb->len;
 471 }
 472 #endif
 473
 474 static const size_t cta_min_ip[CTA_IP_MAX] = {
 475         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 476         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 477 };
 478
 479 static inline int
 480 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 481 {
 482         struct nfattr *tb[CTA_IP_MAX];
 483
 484         DEBUGP("entered %s\n", __FUNCTION__);
 485
 486         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 487
 488         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 489                 return -EINVAL;
 490
 491         if (!tb[CTA_IP_V4_SRC-1])
 492                 return -EINVAL;
 493         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 494
 495         if (!tb[CTA_IP_V4_DST-1])
 496                 return -EINVAL;
 497         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 498
 499         DEBUGP("leaving\n");
 500
 501         return 0;
 502 }
 503
 504 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 505         [CTA_PROTO_NUM-1]       = sizeof(u_int16_t),
 506         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 507         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 508         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 509         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 510         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 511 };
 512
 513 static inline int
 514 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 515                             struct ip_conntrack_tuple *tuple)
 516 {
 517         struct nfattr *tb[CTA_PROTO_MAX];
 518         struct ip_conntrack_protocol *proto;
 519         int ret = 0;
 520
 521         DEBUGP("entered %s\n", __FUNCTION__);
 522
 523         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 524
 525         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 526                 return -EINVAL;
 527
 528         if (!tb[CTA_PROTO_NUM-1])
 529                 return -EINVAL;
 530         tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 531
 532         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 533
 534         if (likely(proto->nfattr_to_tuple))
 535                 ret = proto->nfattr_to_tuple(tb, tuple);
 536
 537         ip_conntrack_proto_put(proto);
 538
 539         return ret;
 540 }
 541
 542 static inline int
 543 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 544                       enum ctattr_tuple type)
 545 {
 546         struct nfattr *tb[CTA_TUPLE_MAX];
 547         int err;
 548
 549         DEBUGP("entered %s\n", __FUNCTION__);
 550
 551         memset(tuple, 0, sizeof(*tuple));
 552
 553         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 554
 555         if (!tb[CTA_TUPLE_IP-1])
 556                 return -EINVAL;
 557
 558         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 559         if (err < 0)
 560                 return err;
 561
 562         if (!tb[CTA_TUPLE_PROTO-1])
 563                 return -EINVAL;
 564
 565         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 566         if (err < 0)
 567                 return err;
 568
 569         /* orig and expect tuples get DIR_ORIGINAL */
 570         if (type == CTA_TUPLE_REPLY)
 571                 tuple->dst.dir = IP_CT_DIR_REPLY;
 572         else
 573                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 574
 575         DUMP_TUPLE(tuple);
 576
 577         DEBUGP("leaving\n");
 578
 579         return 0;
 580 }
 581
 582 #ifdef CONFIG_IP_NF_NAT_NEEDED
 583 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 584         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 585         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 586 };
 587
 588 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 589                                      const struct ip_conntrack *ct,
 590                                      struct ip_nat_range *range)
 591 {
 592         struct nfattr *tb[CTA_PROTONAT_MAX];
 593         struct ip_nat_protocol *npt;
 594
 595         DEBUGP("entered %s\n", __FUNCTION__);
 596
 597         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 598
 599         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 600                 return -EINVAL;
 601
 602         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 603
 604         if (!npt->nfattr_to_range) {
 605                 ip_nat_proto_put(npt);
 606                 return 0;
 607         }
 608
 609         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 610         if (npt->nfattr_to_range(tb, range) > 0)
 611                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 612
 613         ip_nat_proto_put(npt);
 614
 615         DEBUGP("leaving\n");
 616         return 0;
 617 }
 618
 619 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 620         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 621         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 622 };
 623
 624 static inline int
 625 ctnetlink_parse_nat(struct nfattr *cda[],
 626                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 627 {
 628         struct nfattr *tb[CTA_NAT_MAX];
 629         int err;
 630
 631         DEBUGP("entered %s\n", __FUNCTION__);
 632
 633         memset(range, 0, sizeof(*range));
 634
 635         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 636
 637         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 638                 return -EINVAL;
 639
 640         if (tb[CTA_NAT_MINIP-1])
 641                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 642
 643         if (!tb[CTA_NAT_MAXIP-1])
 644                 range->max_ip = range->min_ip;
 645         else
 646                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 647
 648         if (range->min_ip)
 649                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 650
 651         if (!tb[CTA_NAT_PROTO-1])
 652                 return 0;
 653
 654         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 655         if (err < 0)
 656                 return err;
 657
 658         DEBUGP("leaving\n");
 659         return 0;
 660 }
 661 #endif
 662
 663 static inline int
 664 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 665 {
 666         struct nfattr *tb[CTA_HELP_MAX];
 667
 668         DEBUGP("entered %s\n", __FUNCTION__);
 669
 670         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 671
 672         if (!tb[CTA_HELP_NAME-1])
 673                 return -EINVAL;
 674
 675         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 676
 677         return 0;
 678 }
 679
 680 static const size_t cta_min[CTA_MAX] = {
 681         [CTA_STATUS-1]          = sizeof(u_int32_t),
 682         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 683         [CTA_MARK-1]            = sizeof(u_int32_t),
 684         [CTA_USE-1]             = sizeof(u_int32_t),
 685         [CTA_ID-1]              = sizeof(u_int32_t)
 686 };
 687
 688 static int
 689 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 690                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 691 {
 692         struct ip_conntrack_tuple_hash *h;
 693         struct ip_conntrack_tuple tuple;
 694         struct ip_conntrack *ct;
 695         int err = 0;
 696
 697         DEBUGP("entered %s\n", __FUNCTION__);
 698
 699         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 700                 return -EINVAL;
 701
 702         if (cda[CTA_TUPLE_ORIG-1])
 703                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 704         else if (cda[CTA_TUPLE_REPLY-1])
 705                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 706         else {
 707                 /* Flush the whole table */
 708                 ip_conntrack_flush();
 709                 return 0;
 710         }
 711
 712         if (err < 0)
 713                 return err;
 714
 715         h = ip_conntrack_find_get(&tuple, NULL);
 716         if (!h) {
 717                 DEBUGP("tuple not found in conntrack hash\n");
 718                 return -ENOENT;
 719         }
 720
 721         ct = tuplehash_to_ctrack(h);
 722
 723         if (cda[CTA_ID-1]) {
 724                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 725                 if (ct->id != id) {
 726                         ip_conntrack_put(ct);
 727                         return -ENOENT;
 728                 }
 729         }
 730         if (del_timer(&ct->timeout)) {
 731                 ip_conntrack_put(ct);
 732                 ct->timeout.function((unsigned long)ct);
 733                 return 0;
 734         }
 735         ip_conntrack_put(ct);
 736         DEBUGP("leaving\n");
 737
 738         return 0;
 739 }
 740
 741 static int
 742 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 743                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 744 {
 745         struct ip_conntrack_tuple_hash *h;
 746         struct ip_conntrack_tuple tuple;
 747         struct ip_conntrack *ct;
 748         struct sk_buff *skb2 = NULL;
 749         int err = 0;
 750
 751         DEBUGP("entered %s\n", __FUNCTION__);
 752
 753         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 754                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 755                 u32 rlen;
 756
 757                 if (msg->nfgen_family != AF_INET)
 758                         return -EAFNOSUPPORT;
 759
 760                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 761                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 762 #ifdef CONFIG_IP_NF_CT_ACCT
 763                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 764                                                 ctnetlink_dump_table_w,
 765                                                 ctnetlink_done)) != 0)
 766                                 return -EINVAL;
 767 #else
 768                         return -ENOTSUPP;
 769 #endif
 770                 } else {
 771                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 772                                                         ctnetlink_dump_table,
 773                                                         ctnetlink_done)) != 0)
 774                         return -EINVAL;
 775                 }
 776
 777                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 778                 if (rlen > skb->len)
 779                         rlen = skb->len;
 780                 skb_pull(skb, rlen);
 781                 return 0;
 782         }
 783
 784         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 785                 return -EINVAL;
 786
 787         if (cda[CTA_TUPLE_ORIG-1])
 788                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 789         else if (cda[CTA_TUPLE_REPLY-1])
 790                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 791         else
 792                 return -EINVAL;
 793
 794         if (err < 0)
 795                 return err;
 796
 797         h = ip_conntrack_find_get(&tuple, NULL);
 798         if (!h) {
 799                 DEBUGP("tuple not found in conntrack hash");
 800                 return -ENOENT;
 801         }
 802         DEBUGP("tuple found\n");
 803         ct = tuplehash_to_ctrack(h);
 804
 805         err = -ENOMEM;
 806         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 807         if (!skb2) {
 808                 ip_conntrack_put(ct);
 809                 return -ENOMEM;
 810         }
 811         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 812
 813         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 814                                   IPCTNL_MSG_CT_NEW, 1, ct);
 815         ip_conntrack_put(ct);
 816         if (err <= 0)
 817                 goto free;
 818
 819         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 820         if (err < 0)
 821                 goto out;
 822
 823         DEBUGP("leaving\n");
 824         return 0;
 825
 826 free:
 827         kfree_skb(skb2);
 828 out:
 829         return err;
 830 }
 831
 832 static inline int
 833 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 834 {
 835         unsigned long d;
 836         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 837         d = ct->status ^ status;
 838
 839         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 840                 /* unchangeable */
 841                 return -EINVAL;
 842
 843         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 844                 /* SEEN_REPLY bit can only be set */
 845                 return -EINVAL;
 846
 847
 848         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 849                 /* ASSURED bit can only be set */
 850                 return -EINVAL;
 851
 852         if (cda[CTA_NAT-1]) {
 853 #ifndef CONFIG_IP_NF_NAT_NEEDED
 854                 return -EINVAL;
 855 #else
 856                 unsigned int hooknum;
 857                 struct ip_nat_range range;
 858
 859                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 860                         return -EINVAL;
 861
 862                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 863                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 864                        htons(range.min.all), htons(range.max.all));
 865
 866                 /* This is tricky but it works. ip_nat_setup_info needs the
 867                  * hook number as parameter, so let's do the correct
 868                  * conversion and run away */
 869                 if (status & IPS_SRC_NAT_DONE)
 870                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 871                 else if (status & IPS_DST_NAT_DONE)
 872                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 873                 else
 874                         return -EINVAL; /* Missing NAT flags */
 875
 876                 DEBUGP("NAT status: %lu\n",
 877                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 878
 879                 if (ip_nat_initialized(ct, hooknum))
 880                         return -EEXIST;
 881                 ip_nat_setup_info(ct, &range, hooknum);
 882
 883                 DEBUGP("NAT status after setup_info: %lu\n",
 884                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 885 #endif
 886         }
 887
 888         /* Be careful here, modifying NAT bits can screw up things,
 889          * so don't let users modify them directly if they don't pass
 890          * ip_nat_range. */
 891         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 892         return 0;
 893 }
 894
 895
 896 static inline int
 897 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 898 {
 899         struct ip_conntrack_helper *helper;
 900         char *helpname;
 901         int err;
 902
 903         DEBUGP("entered %s\n", __FUNCTION__);
 904
 905         /* don't change helper of sibling connections */
 906         if (ct->master)
 907                 return -EINVAL;
 908
 909         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 910         if (err < 0)
 911                 return err;
 912
 913         helper = __ip_conntrack_helper_find_byname(helpname);
 914         if (!helper) {
 915                 if (!strcmp(helpname, ""))
 916                         helper = NULL;
 917                 else
 918                         return -EINVAL;
 919         }
 920
 921         if (ct->helper) {
 922                 if (!helper) {
 923                         /* we had a helper before ... */
 924                         ip_ct_remove_expectations(ct);
 925                         ct->helper = NULL;
 926                 } else {
 927                         /* need to zero data of old helper */
 928                         memset(&ct->help, 0, sizeof(ct->help));
 929                 }
 930         }
 931
 932         ct->helper = helper;
 933
 934         return 0;
 935 }
 936
 937 static inline int
 938 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 939 {
 940         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 941
 942         if (!del_timer(&ct->timeout))
 943                 return -ETIME;
 944
 945         ct->timeout.expires = jiffies + timeout * HZ;
 946         add_timer(&ct->timeout);
 947
 948         return 0;
 949 }
 950
 951 static inline int
 952 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 953 {
 954         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 955         struct ip_conntrack_protocol *proto;
 956         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 957         int err = 0;
 958
 959         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 960
 961         proto = ip_conntrack_proto_find_get(npt);
 962
 963         if (proto->from_nfattr)
 964                 err = proto->from_nfattr(tb, ct);
 965         ip_conntrack_proto_put(proto);
 966
 967         return err;
 968 }
 969
 970 static int
 971 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 972 {
 973         int err;
 974
 975         DEBUGP("entered %s\n", __FUNCTION__);
 976
 977         if (cda[CTA_HELP-1]) {
 978                 err = ctnetlink_change_helper(ct, cda);
 979                 if (err < 0)
 980                         return err;
 981         }
 982
 983         if (cda[CTA_TIMEOUT-1]) {
 984                 err = ctnetlink_change_timeout(ct, cda);
 985                 if (err < 0)
 986                         return err;
 987         }
 988
 989         if (cda[CTA_STATUS-1]) {
 990                 err = ctnetlink_change_status(ct, cda);
 991                 if (err < 0)
 992                         return err;
 993         }
 994
 995         if (cda[CTA_PROTOINFO-1]) {
 996                 err = ctnetlink_change_protoinfo(ct, cda);
 997                 if (err < 0)
 998                         return err;
 999         }
1000
1001 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1002         if (cda[CTA_MARK-1])
1003                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1004 #endif
1005
1006         DEBUGP("all done\n");
1007         return 0;
1008 }
1009
1010 static int
1011 ctnetlink_create_conntrack(struct nfattr *cda[],
1012                            struct ip_conntrack_tuple *otuple,
1013                            struct ip_conntrack_tuple *rtuple)
1014 {
1015         struct ip_conntrack *ct;
1016         int err = -EINVAL;
1017
1018         DEBUGP("entered %s\n", __FUNCTION__);
1019
1020         ct = ip_conntrack_alloc(otuple, rtuple);
1021         if (ct == NULL || IS_ERR(ct))
1022                 return -ENOMEM;
1023
1024         if (!cda[CTA_TIMEOUT-1])
1025                 goto err;
1026         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1027
1028         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1029         ct->status |= IPS_CONFIRMED;
1030
1031         err = ctnetlink_change_status(ct, cda);
1032         if (err < 0)
1033                 goto err;
1034
1035         if (cda[CTA_PROTOINFO-1]) {
1036                 err = ctnetlink_change_protoinfo(ct, cda);
1037                 if (err < 0)
1038                         return err;
1039         }
1040
1041         ct->helper = ip_conntrack_helper_find_get(rtuple);
1042
1043         add_timer(&ct->timeout);
1044         ip_conntrack_hash_insert(ct);
1045
1046         if (ct->helper)
1047                 ip_conntrack_helper_put(ct->helper);
1048
1049 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1050         if (cda[CTA_MARK-1])
1051                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1052 #endif
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 (nfattr_bad_size(cda, CTA_MAX, cta_min))
1073                 return -EINVAL;
1074
1075         if (cda[CTA_TUPLE_ORIG-1]) {
1076                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1077                 if (err < 0)
1078                         return err;
1079         }
1080
1081         if (cda[CTA_TUPLE_REPLY-1]) {
1082                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1083                 if (err < 0)
1084                         return err;
1085         }
1086
1087         write_lock_bh(&ip_conntrack_lock);
1088         if (cda[CTA_TUPLE_ORIG-1])
1089                 h = __ip_conntrack_find(&otuple, NULL);
1090         else if (cda[CTA_TUPLE_REPLY-1])
1091                 h = __ip_conntrack_find(&rtuple, NULL);
1092
1093         if (h == NULL) {
1094                 write_unlock_bh(&ip_conntrack_lock);
1095                 DEBUGP("no such conntrack, create new\n");
1096                 err = -ENOENT;
1097                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1098                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1099                 return err;
1100         }
1101         /* implicit 'else' */
1102
1103         /* we only allow nat config for new conntracks */
1104         if (cda[CTA_NAT-1]) {
1105                 err = -EINVAL;
1106                 goto out_unlock;
1107         }
1108
1109         /* We manipulate the conntrack inside the global conntrack table lock,
1110          * so there's no need to increase the refcount */
1111         DEBUGP("conntrack found\n");
1112         err = -EEXIST;
1113         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1114                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1115
1116 out_unlock:
1117         write_unlock_bh(&ip_conntrack_lock);
1118         return err;
1119 }
1120
1121 /***********************************************************************
1122  * EXPECT
1123  ***********************************************************************/
1124
1125 static inline int
1126 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1127                          const struct ip_conntrack_tuple *tuple,
1128                          enum ctattr_expect type)
1129 {
1130         struct nfattr *nest_parms = NFA_NEST(skb, type);
1131
1132         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1133                 goto nfattr_failure;
1134
1135         NFA_NEST_END(skb, nest_parms);
1136
1137         return 0;
1138
1139 nfattr_failure:
1140         return -1;
1141 }
1142
1143 static inline int
1144 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1145                           const struct ip_conntrack_expect *exp)
1146 {
1147         struct ip_conntrack *master = exp->master;
1148         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1149         u_int32_t id = htonl(exp->id);
1150
1151         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1152                 goto nfattr_failure;
1153         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1154                 goto nfattr_failure;
1155         if (ctnetlink_exp_dump_tuple(skb,
1156                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1157                                  CTA_EXPECT_MASTER) < 0)
1158                 goto nfattr_failure;
1159
1160         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1161         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1162
1163         return 0;
1164
1165 nfattr_failure:
1166         return -1;
1167 }
1168
1169 static int
1170 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1171                     int event,
1172                     int nowait,
1173                     const struct ip_conntrack_expect *exp)
1174 {
1175         struct nlmsghdr *nlh;
1176         struct nfgenmsg *nfmsg;
1177         unsigned char *b;
1178
1179         b = skb->tail;
1180
1181         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1182         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1183         nfmsg  = NLMSG_DATA(nlh);
1184
1185         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1186         nfmsg->nfgen_family = AF_INET;
1187         nfmsg->version      = NFNETLINK_V0;
1188         nfmsg->res_id       = 0;
1189
1190         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1191                 goto nfattr_failure;
1192
1193         nlh->nlmsg_len = skb->tail - b;
1194         return skb->len;
1195
1196 nlmsg_failure:
1197 nfattr_failure:
1198         skb_trim(skb, b - skb->data);
1199         return -1;
1200 }
1201
1202 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1203 static int ctnetlink_expect_event(struct notifier_block *this,
1204                                   unsigned long events, void *ptr)
1205 {
1206         struct nlmsghdr *nlh;
1207         struct nfgenmsg *nfmsg;
1208         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1209         struct sk_buff *skb;
1210         unsigned int type;
1211         unsigned char *b;
1212         int flags = 0;
1213         u16 proto;
1214
1215         if (events & IPEXP_NEW) {
1216                 type = IPCTNL_MSG_EXP_NEW;
1217                 flags = NLM_F_CREATE|NLM_F_EXCL;
1218         } else
1219                 return NOTIFY_DONE;
1220
1221         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1222         if (!skb)
1223                 return NOTIFY_DONE;
1224
1225         b = skb->tail;
1226
1227         type |= NFNL_SUBSYS_CTNETLINK << 8;
1228         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1229         nfmsg = NLMSG_DATA(nlh);
1230
1231         nlh->nlmsg_flags    = flags;
1232         nfmsg->nfgen_family = AF_INET;
1233         nfmsg->version      = NFNETLINK_V0;
1234         nfmsg->res_id       = 0;
1235
1236         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1237                 goto nfattr_failure;
1238
1239         nlh->nlmsg_len = skb->tail - b;
1240         proto = exp->tuple.dst.protonum;
1241         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1242         return NOTIFY_DONE;
1243
1244 nlmsg_failure:
1245 nfattr_failure:
1246         kfree_skb(skb);
1247         return NOTIFY_DONE;
1248 }
1249 #endif
1250
1251 static int
1252 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1253 {
1254         struct ip_conntrack_expect *exp = NULL;
1255         struct list_head *i;
1256         u_int32_t *id = (u_int32_t *) &cb->args[0];
1257
1258         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1259
1260         read_lock_bh(&ip_conntrack_lock);
1261         list_for_each_prev(i, &ip_conntrack_expect_list) {
1262                 exp = (struct ip_conntrack_expect *) i;
1263                 if (exp->id <= *id)
1264                         continue;
1265                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1266                                             cb->nlh->nlmsg_seq,
1267                                             IPCTNL_MSG_EXP_NEW,
1268                                             1, exp) < 0)
1269                         goto out;
1270                 *id = exp->id;
1271         }
1272 out:
1273         read_unlock_bh(&ip_conntrack_lock);
1274
1275         DEBUGP("leaving, last id=%llu\n", *id);
1276
1277         return skb->len;
1278 }
1279
1280 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1281         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1282         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1283 };
1284
1285 static int
1286 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1287                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1288 {
1289         struct ip_conntrack_tuple tuple;
1290         struct ip_conntrack_expect *exp;
1291         struct sk_buff *skb2;
1292         int err = 0;
1293
1294         DEBUGP("entered %s\n", __FUNCTION__);
1295
1296         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1297                 return -EINVAL;
1298
1299         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1300                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1301                 u32 rlen;
1302
1303                 if (msg->nfgen_family != AF_INET)
1304                         return -EAFNOSUPPORT;
1305
1306                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1307                                                 ctnetlink_exp_dump_table,
1308                                                 ctnetlink_done)) != 0)
1309                         return -EINVAL;
1310                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1311                 if (rlen > skb->len)
1312                         rlen = skb->len;
1313                 skb_pull(skb, rlen);
1314                 return 0;
1315         }
1316
1317         if (cda[CTA_EXPECT_MASTER-1])
1318                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1319         else
1320                 return -EINVAL;
1321
1322         if (err < 0)
1323                 return err;
1324
1325         exp = ip_conntrack_expect_find(&tuple);
1326         if (!exp)
1327                 return -ENOENT;
1328
1329         if (cda[CTA_EXPECT_ID-1]) {
1330                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1331                 if (exp->id != ntohl(id)) {
1332                         ip_conntrack_expect_put(exp);
1333                         return -ENOENT;
1334                 }
1335         }
1336
1337         err = -ENOMEM;
1338         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1339         if (!skb2)
1340                 goto out;
1341         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1342
1343         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1344                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1345                                       1, exp);
1346         if (err <= 0)
1347                 goto free;
1348
1349         ip_conntrack_expect_put(exp);
1350
1351         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1352
1353 free:
1354         kfree_skb(skb2);
1355 out:
1356         ip_conntrack_expect_put(exp);
1357         return err;
1358 }
1359
1360 static int
1361 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1362                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1363 {
1364         struct ip_conntrack_expect *exp, *tmp;
1365         struct ip_conntrack_tuple tuple;
1366         struct ip_conntrack_helper *h;
1367         int err;
1368
1369         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1370                 return -EINVAL;
1371
1372         if (cda[CTA_EXPECT_TUPLE-1]) {
1373                 /* delete a single expect by tuple */
1374                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1375                 if (err < 0)
1376                         return err;
1377
1378                 /* bump usage count to 2 */
1379                 exp = ip_conntrack_expect_find(&tuple);
1380                 if (!exp)
1381                         return -ENOENT;
1382
1383                 if (cda[CTA_EXPECT_ID-1]) {
1384                         u_int32_t id =
1385                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1386                         if (exp->id != ntohl(id)) {
1387                                 ip_conntrack_expect_put(exp);
1388                                 return -ENOENT;
1389                         }
1390                 }
1391
1392                 /* after list removal, usage count == 1 */
1393                 ip_conntrack_unexpect_related(exp);
1394                 /* have to put what we 'get' above.
1395                  * after this line usage count == 0 */
1396                 ip_conntrack_expect_put(exp);
1397         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1398                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1399
1400                 /* delete all expectations for this helper */
1401                 write_lock_bh(&ip_conntrack_lock);
1402                 h = __ip_conntrack_helper_find_byname(name);
1403                 if (!h) {
1404                         write_unlock_bh(&ip_conntrack_lock);
1405                         return -EINVAL;
1406                 }
1407                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1408                                          list) {
1409                         if (exp->master->helper == h
1410                             && del_timer(&exp->timeout)) {
1411                                 ip_ct_unlink_expect(exp);
1412                                 ip_conntrack_expect_put(exp);
1413                         }
1414                 }
1415                 write_unlock_bh(&ip_conntrack_lock);
1416         } else {
1417                 /* This basically means we have to flush everything*/
1418                 write_lock_bh(&ip_conntrack_lock);
1419                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1420                                          list) {
1421                         if (del_timer(&exp->timeout)) {
1422                                 ip_ct_unlink_expect(exp);
1423                                 ip_conntrack_expect_put(exp);
1424                         }
1425                 }
1426                 write_unlock_bh(&ip_conntrack_lock);
1427         }
1428
1429         return 0;
1430 }
1431 static int
1432 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1433 {
1434         return -EOPNOTSUPP;
1435 }
1436
1437 static int
1438 ctnetlink_create_expect(struct nfattr *cda[])
1439 {
1440         struct ip_conntrack_tuple tuple, mask, master_tuple;
1441         struct ip_conntrack_tuple_hash *h = NULL;
1442         struct ip_conntrack_expect *exp;
1443         struct ip_conntrack *ct;
1444         int err = 0;
1445
1446         DEBUGP("entered %s\n", __FUNCTION__);
1447
1448         /* caller guarantees that those three CTA_EXPECT_* exist */
1449         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1450         if (err < 0)
1451                 return err;
1452         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1453         if (err < 0)
1454                 return err;
1455         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1456         if (err < 0)
1457                 return err;
1458
1459         /* Look for master conntrack of this expectation */
1460         h = ip_conntrack_find_get(&master_tuple, NULL);
1461         if (!h)
1462                 return -ENOENT;
1463         ct = tuplehash_to_ctrack(h);
1464
1465         if (!ct->helper) {
1466                 /* such conntrack hasn't got any helper, abort */
1467                 err = -EINVAL;
1468                 goto out;
1469         }
1470
1471         exp = ip_conntrack_expect_alloc(ct);
1472         if (!exp) {
1473                 err = -ENOMEM;
1474                 goto out;
1475         }
1476
1477         exp->expectfn = NULL;
1478         exp->flags = 0;
1479         exp->master = ct;
1480         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1481         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1482
1483         err = ip_conntrack_expect_related(exp);
1484         ip_conntrack_expect_put(exp);
1485
1486 out:
1487         ip_conntrack_put(tuplehash_to_ctrack(h));
1488         return err;
1489 }
1490
1491 static int
1492 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1493                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1494 {
1495         struct ip_conntrack_tuple tuple;
1496         struct ip_conntrack_expect *exp;
1497         int err = 0;
1498
1499         DEBUGP("entered %s\n", __FUNCTION__);
1500
1501         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1502                 return -EINVAL;
1503
1504         if (!cda[CTA_EXPECT_TUPLE-1]
1505             || !cda[CTA_EXPECT_MASK-1]
1506             || !cda[CTA_EXPECT_MASTER-1])
1507                 return -EINVAL;
1508
1509         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1510         if (err < 0)
1511                 return err;
1512
1513         write_lock_bh(&ip_conntrack_lock);
1514         exp = __ip_conntrack_expect_find(&tuple);
1515
1516         if (!exp) {
1517                 write_unlock_bh(&ip_conntrack_lock);
1518                 err = -ENOENT;
1519                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1520                         err = ctnetlink_create_expect(cda);
1521                 return err;
1522         }
1523
1524         err = -EEXIST;
1525         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1526                 err = ctnetlink_change_expect(exp, cda);
1527         write_unlock_bh(&ip_conntrack_lock);
1528
1529         DEBUGP("leaving\n");
1530
1531         return err;
1532 }
1533
1534 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1535 static struct notifier_block ctnl_notifier = {
1536         .notifier_call  = ctnetlink_conntrack_event,
1537 };
1538
1539 static struct notifier_block ctnl_notifier_exp = {
1540         .notifier_call  = ctnetlink_expect_event,
1541 };
1542 #endif
1543
1544 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1545         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1546                                             .attr_count = CTA_MAX, },
1547         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1548                                             .attr_count = CTA_MAX, },
1549         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1550                                             .attr_count = CTA_MAX, },
1551         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1552                                             .attr_count = CTA_MAX, },
1553 };
1554
1555 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1556         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1557                                             .attr_count = CTA_EXPECT_MAX, },
1558         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1559                                             .attr_count = CTA_EXPECT_MAX, },
1560         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1561                                             .attr_count = CTA_EXPECT_MAX, },
1562 };
1563
1564 static struct nfnetlink_subsystem ctnl_subsys = {
1565         .name                           = "conntrack",
1566         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1567         .cb_count                       = IPCTNL_MSG_MAX,
1568         .cb                             = ctnl_cb,
1569 };
1570
1571 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1572         .name                           = "conntrack_expect",
1573         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1574         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1575         .cb                             = ctnl_exp_cb,
1576 };
1577
1578 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1579
1580 static int __init ctnetlink_init(void)
1581 {
1582         int ret;
1583
1584         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1585         ret = nfnetlink_subsys_register(&ctnl_subsys);
1586         if (ret < 0) {
1587                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1588                 goto err_out;
1589         }
1590
1591         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1592         if (ret < 0) {
1593                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1594                 goto err_unreg_subsys;
1595         }
1596
1597 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1598         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1599         if (ret < 0) {
1600                 printk("ctnetlink_init: cannot register notifier.\n");
1601                 goto err_unreg_exp_subsys;
1602         }
1603
1604         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1605         if (ret < 0) {
1606                 printk("ctnetlink_init: cannot expect register notifier.\n");
1607                 goto err_unreg_notifier;
1608         }
1609 #endif
1610
1611         return 0;
1612
1613 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1614 err_unreg_notifier:
1615         ip_conntrack_unregister_notifier(&ctnl_notifier);
1616 err_unreg_exp_subsys:
1617         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1618 #endif
1619 err_unreg_subsys:
1620         nfnetlink_subsys_unregister(&ctnl_subsys);
1621 err_out:
1622         return ret;
1623 }
1624
1625 static void __exit ctnetlink_exit(void)
1626 {
1627         printk("ctnetlink: unregistering from nfnetlink.\n");
1628
1629 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1630         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1631         ip_conntrack_unregister_notifier(&ctnl_notifier);
1632 #endif
1633
1634         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1635         nfnetlink_subsys_unregister(&ctnl_subsys);
1636         return;
1637 }
1638
1639 module_init(ctnetlink_init);
1640 module_exit(ctnetlink_exit);