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) 2005-2006 by Pablo Neira Ayuso <pablo@eurodev.net>
   8  *
   9  * I've reworked this stuff to use attributes instead of conntrack
  10  * structures. 5.44 am. I need more tea. --pablo 05/07/11.
  11  *
  12  * Initial connection tracking via netlink development funded and
  13  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
  14  *
  15  * Further development of this code funded by Astaro AG (http://www.astaro.com)
  16  *
  17  * This software may be used and distributed according to the terms
  18  * of the GNU General Public License, incorporated herein by reference.
  19  */
  20
  21 #include <linux/init.h>
  22 #include <linux/module.h>
  23 #include <linux/kernel.h>
  24 #include <linux/types.h>
  25 #include <linux/timer.h>
  26 #include <linux/skbuff.h>
  27 #include <linux/errno.h>
  28 #include <linux/netlink.h>
  29 #include <linux/spinlock.h>
  30 #include <linux/interrupt.h>
  31 #include <linux/notifier.h>
  32
  33 #include <linux/netfilter.h>
  34 #include <linux/netfilter_ipv4/ip_conntrack.h>
  35 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
  36 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
  37 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
  38 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
  39
  40 #include <linux/netfilter/nfnetlink.h>
  41 #include <linux/netfilter/nfnetlink_conntrack.h>
  42
  43 MODULE_LICENSE("GPL");
  44
  45 static char __initdata version[] = "0.90";
  46
  47 static inline int
  48 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  49                             const struct ip_conntrack_tuple *tuple,
  50                             struct ip_conntrack_protocol *proto)
  51 {
  52         int ret = 0;
  53         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  54
  55         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  56
  57         if (likely(proto->tuple_to_nfattr))
  58                 ret = proto->tuple_to_nfattr(skb, tuple);
  59
  60         NFA_NEST_END(skb, nest_parms);
  61
  62         return ret;
  63
  64 nfattr_failure:
  65         return -1;
  66 }
  67
  68 static inline int
  69 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
  70                          const struct ip_conntrack_tuple *tuple)
  71 {
  72         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  73
  74         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(__be32), &tuple->src.ip);
  75         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(__be32), &tuple->dst.ip);
  76
  77         NFA_NEST_END(skb, nest_parms);
  78
  79         return 0;
  80
  81 nfattr_failure:
  82         return -1;
  83 }
  84
  85 static inline int
  86 ctnetlink_dump_tuples(struct sk_buff *skb,
  87                       const struct ip_conntrack_tuple *tuple)
  88 {
  89         int ret;
  90         struct ip_conntrack_protocol *proto;
  91
  92         ret = ctnetlink_dump_tuples_ip(skb, tuple);
  93         if (unlikely(ret < 0))
  94                 return ret;
  95
  96         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  97         ret = ctnetlink_dump_tuples_proto(skb, tuple, proto);
  98         ip_conntrack_proto_put(proto);
  99
 100         return ret;
 101 }
 102
 103 static inline int
 104 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 105 {
 106         __be32 status = htonl((u_int32_t) ct->status);
 107         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 108         return 0;
 109
 110 nfattr_failure:
 111         return -1;
 112 }
 113
 114 static inline int
 115 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 116 {
 117         long timeout_l = ct->timeout.expires - jiffies;
 118         __be32 timeout;
 119
 120         if (timeout_l < 0)
 121                 timeout = 0;
 122         else
 123                 timeout = htonl(timeout_l / HZ);
 124
 125         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 126         return 0;
 127
 128 nfattr_failure:
 129         return -1;
 130 }
 131
 132 static inline int
 133 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 134 {
 135         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 136
 137         struct nfattr *nest_proto;
 138         int ret;
 139
 140         if (!proto->to_nfattr) {
 141                 ip_conntrack_proto_put(proto);
 142                 return 0;
 143         }
 144
 145         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 146
 147         ret = proto->to_nfattr(skb, nest_proto, ct);
 148
 149         ip_conntrack_proto_put(proto);
 150
 151         NFA_NEST_END(skb, nest_proto);
 152
 153         return ret;
 154
 155 nfattr_failure:
 156         ip_conntrack_proto_put(proto);
 157         return -1;
 158 }
 159
 160 static inline int
 161 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 162 {
 163         struct nfattr *nest_helper;
 164
 165         if (!ct->helper)
 166                 return 0;
 167
 168         nest_helper = NFA_NEST(skb, CTA_HELP);
 169         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 170
 171         if (ct->helper->to_nfattr)
 172                 ct->helper->to_nfattr(skb, ct);
 173
 174         NFA_NEST_END(skb, nest_helper);
 175
 176         return 0;
 177
 178 nfattr_failure:
 179         return -1;
 180 }
 181
 182 #ifdef CONFIG_IP_NF_CT_ACCT
 183 static inline int
 184 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 185                         enum ip_conntrack_dir dir)
 186 {
 187         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 188         struct nfattr *nest_count = NFA_NEST(skb, type);
 189         __be32 tmp;
 190
 191         tmp = htonl(ct->counters[dir].packets);
 192         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(__be32), &tmp);
 193
 194         tmp = htonl(ct->counters[dir].bytes);
 195         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(__be32), &tmp);
 196
 197         NFA_NEST_END(skb, nest_count);
 198
 199         return 0;
 200
 201 nfattr_failure:
 202         return -1;
 203 }
 204 #else
 205 #define ctnetlink_dump_counters(a, b, c) (0)
 206 #endif
 207
 208 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 209 static inline int
 210 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 211 {
 212         __be32 mark = htonl(ct->mark);
 213
 214         NFA_PUT(skb, CTA_MARK, sizeof(__be32), &mark);
 215         return 0;
 216
 217 nfattr_failure:
 218         return -1;
 219 }
 220 #else
 221 #define ctnetlink_dump_mark(a, b) (0)
 222 #endif
 223
 224 static inline int
 225 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 226 {
 227         __be32 id = htonl(ct->id);
 228         NFA_PUT(skb, CTA_ID, sizeof(__be32), &id);
 229         return 0;
 230
 231 nfattr_failure:
 232         return -1;
 233 }
 234
 235 static inline int
 236 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 237 {
 238         __be32 use = htonl(atomic_read(&ct->ct_general.use));
 239
 240         NFA_PUT(skb, CTA_USE, sizeof(__be32), &use);
 241         return 0;
 242
 243 nfattr_failure:
 244         return -1;
 245 }
 246
 247 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 248
 249 static int
 250 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 251                     int event, int nowait,
 252                     const struct ip_conntrack *ct)
 253 {
 254         struct nlmsghdr *nlh;
 255         struct nfgenmsg *nfmsg;
 256         struct nfattr *nest_parms;
 257         unsigned char *b;
 258
 259         b = skb->tail;
 260
 261         event |= NFNL_SUBSYS_CTNETLINK << 8;
 262         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 263         nfmsg  = NLMSG_DATA(nlh);
 264
 265         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 266         nfmsg->nfgen_family = AF_INET;
 267         nfmsg->version      = NFNETLINK_V0;
 268         nfmsg->res_id       = 0;
 269
 270         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 271         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 272                 goto nfattr_failure;
 273         NFA_NEST_END(skb, nest_parms);
 274
 275         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 276         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 277                 goto nfattr_failure;
 278         NFA_NEST_END(skb, nest_parms);
 279
 280         if (ctnetlink_dump_status(skb, ct) < 0 ||
 281             ctnetlink_dump_timeout(skb, ct) < 0 ||
 282             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 283             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 284             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 285             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 286             ctnetlink_dump_mark(skb, ct) < 0 ||
 287             ctnetlink_dump_id(skb, ct) < 0 ||
 288             ctnetlink_dump_use(skb, ct) < 0)
 289                 goto nfattr_failure;
 290
 291         nlh->nlmsg_len = skb->tail - b;
 292         return skb->len;
 293
 294 nlmsg_failure:
 295 nfattr_failure:
 296         skb_trim(skb, b - skb->data);
 297         return -1;
 298 }
 299
 300 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 301 static int ctnetlink_conntrack_event(struct notifier_block *this,
 302                                      unsigned long events, void *ptr)
 303 {
 304         struct nlmsghdr *nlh;
 305         struct nfgenmsg *nfmsg;
 306         struct nfattr *nest_parms;
 307         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 308         struct sk_buff *skb;
 309         unsigned int type;
 310         unsigned char *b;
 311         unsigned int flags = 0, group;
 312
 313         /* ignore our fake conntrack entry */
 314         if (ct == &ip_conntrack_untracked)
 315                 return NOTIFY_DONE;
 316
 317         if (events & IPCT_DESTROY) {
 318                 type = IPCTNL_MSG_CT_DELETE;
 319                 group = NFNLGRP_CONNTRACK_DESTROY;
 320         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 321                 type = IPCTNL_MSG_CT_NEW;
 322                 flags = NLM_F_CREATE|NLM_F_EXCL;
 323                 /* dump everything */
 324                 events = ~0UL;
 325                 group = NFNLGRP_CONNTRACK_NEW;
 326         } else if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
 327                 type = IPCTNL_MSG_CT_NEW;
 328                 group = NFNLGRP_CONNTRACK_UPDATE;
 329         } else
 330                 return NOTIFY_DONE;
 331
 332         if (!nfnetlink_has_listeners(group))
 333                 return NOTIFY_DONE;
 334
 335         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 336         if (!skb)
 337                 return NOTIFY_DONE;
 338
 339         b = skb->tail;
 340
 341         type |= NFNL_SUBSYS_CTNETLINK << 8;
 342         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 343         nfmsg = NLMSG_DATA(nlh);
 344
 345         nlh->nlmsg_flags    = flags;
 346         nfmsg->nfgen_family = AF_INET;
 347         nfmsg->version  = NFNETLINK_V0;
 348         nfmsg->res_id   = 0;
 349
 350         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 351         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 352                 goto nfattr_failure;
 353         NFA_NEST_END(skb, nest_parms);
 354
 355         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 356         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 357                 goto nfattr_failure;
 358         NFA_NEST_END(skb, nest_parms);
 359
 360         /* NAT stuff is now a status flag */
 361         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 362             && ctnetlink_dump_status(skb, ct) < 0)
 363                 goto nfattr_failure;
 364         if (events & IPCT_REFRESH
 365             && ctnetlink_dump_timeout(skb, ct) < 0)
 366                 goto nfattr_failure;
 367         if (events & IPCT_PROTOINFO
 368             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 369                 goto nfattr_failure;
 370         if (events & IPCT_HELPINFO
 371             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 372                 goto nfattr_failure;
 373
 374         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 375             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 376                 goto nfattr_failure;
 377
 378         if (events & IPCT_MARK
 379             && ctnetlink_dump_mark(skb, ct) < 0)
 380                 goto nfattr_failure;
 381
 382         nlh->nlmsg_len = skb->tail - b;
 383         nfnetlink_send(skb, 0, group, 0);
 384         return NOTIFY_DONE;
 385
 386 nlmsg_failure:
 387 nfattr_failure:
 388         kfree_skb(skb);
 389         return NOTIFY_DONE;
 390 }
 391 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 392
 393 static int ctnetlink_done(struct netlink_callback *cb)
 394 {
 395         if (cb->args[1])
 396                 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
 397         return 0;
 398 }
 399
 400 static int
 401 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 402 {
 403         struct ip_conntrack *ct, *last;
 404         struct ip_conntrack_tuple_hash *h;
 405         struct list_head *i;
 406
 407         read_lock_bh(&ip_conntrack_lock);
 408         last = (struct ip_conntrack *)cb->args[1];
 409         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
 410 restart:
 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 (cb->args[1]) {
 417                                 if (ct != last)
 418                                         continue;
 419                                 cb->args[1] = 0;
 420                         }
 421                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 422                                                 cb->nlh->nlmsg_seq,
 423                                                 IPCTNL_MSG_CT_NEW,
 424                                                 1, ct) < 0) {
 425                                 nf_conntrack_get(&ct->ct_general);
 426                                 cb->args[1] = (unsigned long)ct;
 427                                 goto out;
 428                         }
 429 #ifdef CONFIG_NF_CT_ACCT
 430                         if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
 431                                                 IPCTNL_MSG_CT_GET_CTRZERO)
 432                                 memset(&ct->counters, 0, sizeof(ct->counters));
 433 #endif
 434                 }
 435                 if (cb->args[1]) {
 436                         cb->args[1] = 0;
 437                         goto restart;
 438                 }
 439         }
 440 out:
 441         read_unlock_bh(&ip_conntrack_lock);
 442         if (last)
 443                 ip_conntrack_put(last);
 444
 445         return skb->len;
 446 }
 447
 448 static const size_t cta_min_ip[CTA_IP_MAX] = {
 449         [CTA_IP_V4_SRC-1]       = sizeof(__be32),
 450         [CTA_IP_V4_DST-1]       = sizeof(__be32),
 451 };
 452
 453 static inline int
 454 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 455 {
 456         struct nfattr *tb[CTA_IP_MAX];
 457
 458         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 459
 460         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 461                 return -EINVAL;
 462
 463         if (!tb[CTA_IP_V4_SRC-1])
 464                 return -EINVAL;
 465         tuple->src.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 466
 467         if (!tb[CTA_IP_V4_DST-1])
 468                 return -EINVAL;
 469         tuple->dst.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 470
 471         return 0;
 472 }
 473
 474 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 475         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 476         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 477         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 478         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 479         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 480         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 481 };
 482
 483 static inline int
 484 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 485                             struct ip_conntrack_tuple *tuple)
 486 {
 487         struct nfattr *tb[CTA_PROTO_MAX];
 488         struct ip_conntrack_protocol *proto;
 489         int ret = 0;
 490
 491         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 492
 493         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 494                 return -EINVAL;
 495
 496         if (!tb[CTA_PROTO_NUM-1])
 497                 return -EINVAL;
 498         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 499
 500         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 501
 502         if (likely(proto->nfattr_to_tuple))
 503                 ret = proto->nfattr_to_tuple(tb, tuple);
 504
 505         ip_conntrack_proto_put(proto);
 506
 507         return ret;
 508 }
 509
 510 static inline int
 511 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 512                       enum ctattr_tuple type)
 513 {
 514         struct nfattr *tb[CTA_TUPLE_MAX];
 515         int err;
 516
 517         memset(tuple, 0, sizeof(*tuple));
 518
 519         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 520
 521         if (!tb[CTA_TUPLE_IP-1])
 522                 return -EINVAL;
 523
 524         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 525         if (err < 0)
 526                 return err;
 527
 528         if (!tb[CTA_TUPLE_PROTO-1])
 529                 return -EINVAL;
 530
 531         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 532         if (err < 0)
 533                 return err;
 534
 535         /* orig and expect tuples get DIR_ORIGINAL */
 536         if (type == CTA_TUPLE_REPLY)
 537                 tuple->dst.dir = IP_CT_DIR_REPLY;
 538         else
 539                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 540
 541         return 0;
 542 }
 543
 544 #ifdef CONFIG_IP_NF_NAT_NEEDED
 545 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 546         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 547         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 548 };
 549
 550 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 551                                      const struct ip_conntrack *ct,
 552                                      struct ip_nat_range *range)
 553 {
 554         struct nfattr *tb[CTA_PROTONAT_MAX];
 555         struct ip_nat_protocol *npt;
 556
 557         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 558
 559         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 560                 return -EINVAL;
 561
 562         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 563
 564         if (!npt->nfattr_to_range) {
 565                 ip_nat_proto_put(npt);
 566                 return 0;
 567         }
 568
 569         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 570         if (npt->nfattr_to_range(tb, range) > 0)
 571                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 572
 573         ip_nat_proto_put(npt);
 574
 575         return 0;
 576 }
 577
 578 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 579         [CTA_NAT_MINIP-1]       = sizeof(__be32),
 580         [CTA_NAT_MAXIP-1]       = sizeof(__be32),
 581 };
 582
 583 static inline int
 584 ctnetlink_parse_nat(struct nfattr *nat,
 585                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 586 {
 587         struct nfattr *tb[CTA_NAT_MAX];
 588         int err;
 589
 590         memset(range, 0, sizeof(*range));
 591
 592         nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
 593
 594         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 595                 return -EINVAL;
 596
 597         if (tb[CTA_NAT_MINIP-1])
 598                 range->min_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 599
 600         if (!tb[CTA_NAT_MAXIP-1])
 601                 range->max_ip = range->min_ip;
 602         else
 603                 range->max_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 604
 605         if (range->min_ip)
 606                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 607
 608         if (!tb[CTA_NAT_PROTO-1])
 609                 return 0;
 610
 611         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 612         if (err < 0)
 613                 return err;
 614
 615         return 0;
 616 }
 617 #endif
 618
 619 static inline int
 620 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 621 {
 622         struct nfattr *tb[CTA_HELP_MAX];
 623
 624         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 625
 626         if (!tb[CTA_HELP_NAME-1])
 627                 return -EINVAL;
 628
 629         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 630
 631         return 0;
 632 }
 633
 634 static const size_t cta_min[CTA_MAX] = {
 635         [CTA_STATUS-1]          = sizeof(__be32),
 636         [CTA_TIMEOUT-1]         = sizeof(__be32),
 637         [CTA_MARK-1]            = sizeof(__be32),
 638         [CTA_USE-1]             = sizeof(__be32),
 639         [CTA_ID-1]              = sizeof(__be32)
 640 };
 641
 642 static int
 643 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 644                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 645 {
 646         struct ip_conntrack_tuple_hash *h;
 647         struct ip_conntrack_tuple tuple;
 648         struct ip_conntrack *ct;
 649         int err = 0;
 650
 651         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 652                 return -EINVAL;
 653
 654         if (cda[CTA_TUPLE_ORIG-1])
 655                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 656         else if (cda[CTA_TUPLE_REPLY-1])
 657                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 658         else {
 659                 /* Flush the whole table */
 660                 ip_conntrack_flush();
 661                 return 0;
 662         }
 663
 664         if (err < 0)
 665                 return err;
 666
 667         h = ip_conntrack_find_get(&tuple, NULL);
 668         if (!h)
 669                 return -ENOENT;
 670
 671         ct = tuplehash_to_ctrack(h);
 672
 673         if (cda[CTA_ID-1]) {
 674                 u_int32_t id = ntohl(*(__be32 *)NFA_DATA(cda[CTA_ID-1]));
 675                 if (ct->id != id) {
 676                         ip_conntrack_put(ct);
 677                         return -ENOENT;
 678                 }
 679         }
 680         if (del_timer(&ct->timeout))
 681                 ct->timeout.function((unsigned long)ct);
 682
 683         ip_conntrack_put(ct);
 684
 685         return 0;
 686 }
 687
 688 static int
 689 ctnetlink_get_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         struct sk_buff *skb2 = NULL;
 696         int err = 0;
 697
 698         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 699                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 700                 u32 rlen;
 701
 702                 if (msg->nfgen_family != AF_INET)
 703                         return -EAFNOSUPPORT;
 704
 705 #ifndef CONFIG_IP_NF_CT_ACCT
 706                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
 707                         return -ENOTSUPP;
 708 #endif
 709                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 710                                                 ctnetlink_dump_table,
 711                                                 ctnetlink_done)) != 0)
 712                         return -EINVAL;
 713
 714                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 715                 if (rlen > skb->len)
 716                         rlen = skb->len;
 717                 skb_pull(skb, rlen);
 718                 return 0;
 719         }
 720
 721         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 722                 return -EINVAL;
 723
 724         if (cda[CTA_TUPLE_ORIG-1])
 725                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 726         else if (cda[CTA_TUPLE_REPLY-1])
 727                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 728         else
 729                 return -EINVAL;
 730
 731         if (err < 0)
 732                 return err;
 733
 734         h = ip_conntrack_find_get(&tuple, NULL);
 735         if (!h)
 736                 return -ENOENT;
 737
 738         ct = tuplehash_to_ctrack(h);
 739
 740         err = -ENOMEM;
 741         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 742         if (!skb2) {
 743                 ip_conntrack_put(ct);
 744                 return -ENOMEM;
 745         }
 746         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 747
 748         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 749                                   IPCTNL_MSG_CT_NEW, 1, ct);
 750         ip_conntrack_put(ct);
 751         if (err <= 0)
 752                 goto free;
 753
 754         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 755         if (err < 0)
 756                 goto out;
 757
 758         return 0;
 759
 760 free:
 761         kfree_skb(skb2);
 762 out:
 763         return err;
 764 }
 765
 766 static inline int
 767 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 768 {
 769         unsigned long d;
 770         unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
 771         d = ct->status ^ status;
 772
 773         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 774                 /* unchangeable */
 775                 return -EINVAL;
 776
 777         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 778                 /* SEEN_REPLY bit can only be set */
 779                 return -EINVAL;
 780
 781
 782         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 783                 /* ASSURED bit can only be set */
 784                 return -EINVAL;
 785
 786         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 787 #ifndef CONFIG_IP_NF_NAT_NEEDED
 788                 return -EINVAL;
 789 #else
 790                 struct ip_nat_range range;
 791
 792                 if (cda[CTA_NAT_DST-1]) {
 793                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 794                                                 &range) < 0)
 795                                 return -EINVAL;
 796                         if (ip_nat_initialized(ct,
 797                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 798                                 return -EEXIST;
 799                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 800                 }
 801                 if (cda[CTA_NAT_SRC-1]) {
 802                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 803                                                 &range) < 0)
 804                                 return -EINVAL;
 805                         if (ip_nat_initialized(ct,
 806                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 807                                 return -EEXIST;
 808                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 809                 }
 810 #endif
 811         }
 812
 813         /* Be careful here, modifying NAT bits can screw up things,
 814          * so don't let users modify them directly if they don't pass
 815          * ip_nat_range. */
 816         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 817         return 0;
 818 }
 819
 820
 821 static inline int
 822 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 823 {
 824         struct ip_conntrack_helper *helper;
 825         char *helpname;
 826         int err;
 827
 828         /* don't change helper of sibling connections */
 829         if (ct->master)
 830                 return -EINVAL;
 831
 832         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 833         if (err < 0)
 834                 return err;
 835
 836         helper = __ip_conntrack_helper_find_byname(helpname);
 837         if (!helper) {
 838                 if (!strcmp(helpname, ""))
 839                         helper = NULL;
 840                 else
 841                         return -EINVAL;
 842         }
 843
 844         if (ct->helper) {
 845                 if (!helper) {
 846                         /* we had a helper before ... */
 847                         ip_ct_remove_expectations(ct);
 848                         ct->helper = NULL;
 849                 } else {
 850                         /* need to zero data of old helper */
 851                         memset(&ct->help, 0, sizeof(ct->help));
 852                 }
 853         }
 854
 855         ct->helper = helper;
 856
 857         return 0;
 858 }
 859
 860 static inline int
 861 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 862 {
 863         u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 864
 865         if (!del_timer(&ct->timeout))
 866                 return -ETIME;
 867
 868         ct->timeout.expires = jiffies + timeout * HZ;
 869         add_timer(&ct->timeout);
 870
 871         return 0;
 872 }
 873
 874 static inline int
 875 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 876 {
 877         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 878         struct ip_conntrack_protocol *proto;
 879         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 880         int err = 0;
 881
 882         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 883
 884         proto = ip_conntrack_proto_find_get(npt);
 885
 886         if (proto->from_nfattr)
 887                 err = proto->from_nfattr(tb, ct);
 888         ip_conntrack_proto_put(proto);
 889
 890         return err;
 891 }
 892
 893 static int
 894 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 895 {
 896         int err;
 897
 898         if (cda[CTA_HELP-1]) {
 899                 err = ctnetlink_change_helper(ct, cda);
 900                 if (err < 0)
 901                         return err;
 902         }
 903
 904         if (cda[CTA_TIMEOUT-1]) {
 905                 err = ctnetlink_change_timeout(ct, cda);
 906                 if (err < 0)
 907                         return err;
 908         }
 909
 910         if (cda[CTA_STATUS-1]) {
 911                 err = ctnetlink_change_status(ct, cda);
 912                 if (err < 0)
 913                         return err;
 914         }
 915
 916         if (cda[CTA_PROTOINFO-1]) {
 917                 err = ctnetlink_change_protoinfo(ct, cda);
 918                 if (err < 0)
 919                         return err;
 920         }
 921
 922 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 923         if (cda[CTA_MARK-1])
 924                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 925 #endif
 926
 927         return 0;
 928 }
 929
 930 static int
 931 ctnetlink_create_conntrack(struct nfattr *cda[],
 932                            struct ip_conntrack_tuple *otuple,
 933                            struct ip_conntrack_tuple *rtuple)
 934 {
 935         struct ip_conntrack *ct;
 936         int err = -EINVAL;
 937
 938         ct = ip_conntrack_alloc(otuple, rtuple);
 939         if (ct == NULL || IS_ERR(ct))
 940                 return -ENOMEM;
 941
 942         if (!cda[CTA_TIMEOUT-1])
 943                 goto err;
 944         ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 945
 946         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
 947         ct->status |= IPS_CONFIRMED;
 948
 949         err = ctnetlink_change_status(ct, cda);
 950         if (err < 0)
 951                 goto err;
 952
 953         if (cda[CTA_PROTOINFO-1]) {
 954                 err = ctnetlink_change_protoinfo(ct, cda);
 955                 if (err < 0)
 956                         return err;
 957         }
 958
 959 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 960         if (cda[CTA_MARK-1])
 961                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 962 #endif
 963
 964         ct->helper = ip_conntrack_helper_find_get(rtuple);
 965
 966         add_timer(&ct->timeout);
 967         ip_conntrack_hash_insert(ct);
 968
 969         if (ct->helper)
 970                 ip_conntrack_helper_put(ct->helper);
 971
 972         return 0;
 973
 974 err:
 975         ip_conntrack_free(ct);
 976         return err;
 977 }
 978
 979 static int
 980 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 981                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 982 {
 983         struct ip_conntrack_tuple otuple, rtuple;
 984         struct ip_conntrack_tuple_hash *h = NULL;
 985         int err = 0;
 986
 987         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 988                 return -EINVAL;
 989
 990         if (cda[CTA_TUPLE_ORIG-1]) {
 991                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
 992                 if (err < 0)
 993                         return err;
 994         }
 995
 996         if (cda[CTA_TUPLE_REPLY-1]) {
 997                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
 998                 if (err < 0)
 999                         return err;
1000         }
1001
1002         write_lock_bh(&ip_conntrack_lock);
1003         if (cda[CTA_TUPLE_ORIG-1])
1004                 h = __ip_conntrack_find(&otuple, NULL);
1005         else if (cda[CTA_TUPLE_REPLY-1])
1006                 h = __ip_conntrack_find(&rtuple, NULL);
1007
1008         if (h == NULL) {
1009                 write_unlock_bh(&ip_conntrack_lock);
1010                 err = -ENOENT;
1011                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1012                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1013                 return err;
1014         }
1015         /* implicit 'else' */
1016
1017         /* we only allow nat config for new conntracks */
1018         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1019                 err = -EINVAL;
1020                 goto out_unlock;
1021         }
1022
1023         /* We manipulate the conntrack inside the global conntrack table lock,
1024          * so there's no need to increase the refcount */
1025         err = -EEXIST;
1026         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1027                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1028
1029 out_unlock:
1030         write_unlock_bh(&ip_conntrack_lock);
1031         return err;
1032 }
1033
1034 /***********************************************************************
1035  * EXPECT
1036  ***********************************************************************/
1037
1038 static inline int
1039 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1040                          const struct ip_conntrack_tuple *tuple,
1041                          enum ctattr_expect type)
1042 {
1043         struct nfattr *nest_parms = NFA_NEST(skb, type);
1044
1045         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1046                 goto nfattr_failure;
1047
1048         NFA_NEST_END(skb, nest_parms);
1049
1050         return 0;
1051
1052 nfattr_failure:
1053         return -1;
1054 }
1055
1056 static inline int
1057 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1058                         const struct ip_conntrack_tuple *tuple,
1059                         const struct ip_conntrack_tuple *mask)
1060 {
1061         int ret;
1062         struct ip_conntrack_protocol *proto;
1063         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1064
1065         ret = ctnetlink_dump_tuples_ip(skb, mask);
1066         if (unlikely(ret < 0))
1067                 goto nfattr_failure;
1068
1069         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1070         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1071         ip_conntrack_proto_put(proto);
1072         if (unlikely(ret < 0))
1073                 goto nfattr_failure;
1074
1075         NFA_NEST_END(skb, nest_parms);
1076
1077         return 0;
1078
1079 nfattr_failure:
1080         return -1;
1081 }
1082
1083 static inline int
1084 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1085                           const struct ip_conntrack_expect *exp)
1086 {
1087         struct ip_conntrack *master = exp->master;
1088         __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1089         __be32 id = htonl(exp->id);
1090
1091         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1092                 goto nfattr_failure;
1093         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1094                 goto nfattr_failure;
1095         if (ctnetlink_exp_dump_tuple(skb,
1096                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1097                                  CTA_EXPECT_MASTER) < 0)
1098                 goto nfattr_failure;
1099
1100         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1101         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1102
1103         return 0;
1104
1105 nfattr_failure:
1106         return -1;
1107 }
1108
1109 static int
1110 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1111                     int event,
1112                     int nowait,
1113                     const struct ip_conntrack_expect *exp)
1114 {
1115         struct nlmsghdr *nlh;
1116         struct nfgenmsg *nfmsg;
1117         unsigned char *b;
1118
1119         b = skb->tail;
1120
1121         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1122         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1123         nfmsg  = NLMSG_DATA(nlh);
1124
1125         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1126         nfmsg->nfgen_family = AF_INET;
1127         nfmsg->version      = NFNETLINK_V0;
1128         nfmsg->res_id       = 0;
1129
1130         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1131                 goto nfattr_failure;
1132
1133         nlh->nlmsg_len = skb->tail - b;
1134         return skb->len;
1135
1136 nlmsg_failure:
1137 nfattr_failure:
1138         skb_trim(skb, b - skb->data);
1139         return -1;
1140 }
1141
1142 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1143 static int ctnetlink_expect_event(struct notifier_block *this,
1144                                   unsigned long events, void *ptr)
1145 {
1146         struct nlmsghdr *nlh;
1147         struct nfgenmsg *nfmsg;
1148         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1149         struct sk_buff *skb;
1150         unsigned int type;
1151         unsigned char *b;
1152         int flags = 0;
1153
1154         if (events & IPEXP_NEW) {
1155                 type = IPCTNL_MSG_EXP_NEW;
1156                 flags = NLM_F_CREATE|NLM_F_EXCL;
1157         } else
1158                 return NOTIFY_DONE;
1159
1160         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1161                 return NOTIFY_DONE;
1162
1163         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1164         if (!skb)
1165                 return NOTIFY_DONE;
1166
1167         b = skb->tail;
1168
1169         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1170         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1171         nfmsg = NLMSG_DATA(nlh);
1172
1173         nlh->nlmsg_flags    = flags;
1174         nfmsg->nfgen_family = AF_INET;
1175         nfmsg->version      = NFNETLINK_V0;
1176         nfmsg->res_id       = 0;
1177
1178         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1179                 goto nfattr_failure;
1180
1181         nlh->nlmsg_len = skb->tail - b;
1182         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1183         return NOTIFY_DONE;
1184
1185 nlmsg_failure:
1186 nfattr_failure:
1187         kfree_skb(skb);
1188         return NOTIFY_DONE;
1189 }
1190 #endif
1191
1192 static int
1193 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1194 {
1195         struct ip_conntrack_expect *exp = NULL;
1196         struct list_head *i;
1197         u_int32_t *id = (u_int32_t *) &cb->args[0];
1198
1199         read_lock_bh(&ip_conntrack_lock);
1200         list_for_each_prev(i, &ip_conntrack_expect_list) {
1201                 exp = (struct ip_conntrack_expect *) i;
1202                 if (exp->id <= *id)
1203                         continue;
1204                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1205                                             cb->nlh->nlmsg_seq,
1206                                             IPCTNL_MSG_EXP_NEW,
1207                                             1, exp) < 0)
1208                         goto out;
1209                 *id = exp->id;
1210         }
1211 out:
1212         read_unlock_bh(&ip_conntrack_lock);
1213
1214         return skb->len;
1215 }
1216
1217 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1218         [CTA_EXPECT_TIMEOUT-1]          = sizeof(__be32),
1219         [CTA_EXPECT_ID-1]               = sizeof(__be32)
1220 };
1221
1222 static int
1223 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1224                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1225 {
1226         struct ip_conntrack_tuple tuple;
1227         struct ip_conntrack_expect *exp;
1228         struct sk_buff *skb2;
1229         int err = 0;
1230
1231         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1232                 return -EINVAL;
1233
1234         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1235                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1236                 u32 rlen;
1237
1238                 if (msg->nfgen_family != AF_INET)
1239                         return -EAFNOSUPPORT;
1240
1241                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1242                                                 ctnetlink_exp_dump_table,
1243                                                 ctnetlink_done)) != 0)
1244                         return -EINVAL;
1245                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1246                 if (rlen > skb->len)
1247                         rlen = skb->len;
1248                 skb_pull(skb, rlen);
1249                 return 0;
1250         }
1251
1252         if (cda[CTA_EXPECT_MASTER-1])
1253                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1254         else
1255                 return -EINVAL;
1256
1257         if (err < 0)
1258                 return err;
1259
1260         exp = ip_conntrack_expect_find(&tuple);
1261         if (!exp)
1262                 return -ENOENT;
1263
1264         if (cda[CTA_EXPECT_ID-1]) {
1265                 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1266                 if (exp->id != ntohl(id)) {
1267                         ip_conntrack_expect_put(exp);
1268                         return -ENOENT;
1269                 }
1270         }
1271
1272         err = -ENOMEM;
1273         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1274         if (!skb2)
1275                 goto out;
1276         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1277
1278         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1279                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1280                                       1, exp);
1281         if (err <= 0)
1282                 goto free;
1283
1284         ip_conntrack_expect_put(exp);
1285
1286         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1287
1288 free:
1289         kfree_skb(skb2);
1290 out:
1291         ip_conntrack_expect_put(exp);
1292         return err;
1293 }
1294
1295 static int
1296 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1297                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1298 {
1299         struct ip_conntrack_expect *exp, *tmp;
1300         struct ip_conntrack_tuple tuple;
1301         struct ip_conntrack_helper *h;
1302         int err;
1303
1304         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1305                 return -EINVAL;
1306
1307         if (cda[CTA_EXPECT_TUPLE-1]) {
1308                 /* delete a single expect by tuple */
1309                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1310                 if (err < 0)
1311                         return err;
1312
1313                 /* bump usage count to 2 */
1314                 exp = ip_conntrack_expect_find(&tuple);
1315                 if (!exp)
1316                         return -ENOENT;
1317
1318                 if (cda[CTA_EXPECT_ID-1]) {
1319                         __be32 id =
1320                                 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1321                         if (exp->id != ntohl(id)) {
1322                                 ip_conntrack_expect_put(exp);
1323                                 return -ENOENT;
1324                         }
1325                 }
1326
1327                 /* after list removal, usage count == 1 */
1328                 ip_conntrack_unexpect_related(exp);
1329                 /* have to put what we 'get' above.
1330                  * after this line usage count == 0 */
1331                 ip_conntrack_expect_put(exp);
1332         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1333                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1334
1335                 /* delete all expectations for this helper */
1336                 write_lock_bh(&ip_conntrack_lock);
1337                 h = __ip_conntrack_helper_find_byname(name);
1338                 if (!h) {
1339                         write_unlock_bh(&ip_conntrack_lock);
1340                         return -EINVAL;
1341                 }
1342                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1343                                          list) {
1344                         if (exp->master->helper == h
1345                             && del_timer(&exp->timeout)) {
1346                                 ip_ct_unlink_expect(exp);
1347                                 ip_conntrack_expect_put(exp);
1348                         }
1349                 }
1350                 write_unlock_bh(&ip_conntrack_lock);
1351         } else {
1352                 /* This basically means we have to flush everything*/
1353                 write_lock_bh(&ip_conntrack_lock);
1354                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1355                                          list) {
1356                         if (del_timer(&exp->timeout)) {
1357                                 ip_ct_unlink_expect(exp);
1358                                 ip_conntrack_expect_put(exp);
1359                         }
1360                 }
1361                 write_unlock_bh(&ip_conntrack_lock);
1362         }
1363
1364         return 0;
1365 }
1366 static int
1367 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1368 {
1369         return -EOPNOTSUPP;
1370 }
1371
1372 static int
1373 ctnetlink_create_expect(struct nfattr *cda[])
1374 {
1375         struct ip_conntrack_tuple tuple, mask, master_tuple;
1376         struct ip_conntrack_tuple_hash *h = NULL;
1377         struct ip_conntrack_expect *exp;
1378         struct ip_conntrack *ct;
1379         int err = 0;
1380
1381         /* caller guarantees that those three CTA_EXPECT_* exist */
1382         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1383         if (err < 0)
1384                 return err;
1385         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1386         if (err < 0)
1387                 return err;
1388         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1389         if (err < 0)
1390                 return err;
1391
1392         /* Look for master conntrack of this expectation */
1393         h = ip_conntrack_find_get(&master_tuple, NULL);
1394         if (!h)
1395                 return -ENOENT;
1396         ct = tuplehash_to_ctrack(h);
1397
1398         if (!ct->helper) {
1399                 /* such conntrack hasn't got any helper, abort */
1400                 err = -EINVAL;
1401                 goto out;
1402         }
1403
1404         exp = ip_conntrack_expect_alloc(ct);
1405         if (!exp) {
1406                 err = -ENOMEM;
1407                 goto out;
1408         }
1409
1410         exp->expectfn = NULL;
1411         exp->flags = 0;
1412         exp->master = ct;
1413         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1414         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1415
1416         err = ip_conntrack_expect_related(exp);
1417         ip_conntrack_expect_put(exp);
1418
1419 out:
1420         ip_conntrack_put(tuplehash_to_ctrack(h));
1421         return err;
1422 }
1423
1424 static int
1425 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1426                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1427 {
1428         struct ip_conntrack_tuple tuple;
1429         struct ip_conntrack_expect *exp;
1430         int err = 0;
1431
1432         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1433                 return -EINVAL;
1434
1435         if (!cda[CTA_EXPECT_TUPLE-1]
1436             || !cda[CTA_EXPECT_MASK-1]
1437             || !cda[CTA_EXPECT_MASTER-1])
1438                 return -EINVAL;
1439
1440         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1441         if (err < 0)
1442                 return err;
1443
1444         write_lock_bh(&ip_conntrack_lock);
1445         exp = __ip_conntrack_expect_find(&tuple);
1446
1447         if (!exp) {
1448                 write_unlock_bh(&ip_conntrack_lock);
1449                 err = -ENOENT;
1450                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1451                         err = ctnetlink_create_expect(cda);
1452                 return err;
1453         }
1454
1455         err = -EEXIST;
1456         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1457                 err = ctnetlink_change_expect(exp, cda);
1458         write_unlock_bh(&ip_conntrack_lock);
1459
1460         return err;
1461 }
1462
1463 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1464 static struct notifier_block ctnl_notifier = {
1465         .notifier_call  = ctnetlink_conntrack_event,
1466 };
1467
1468 static struct notifier_block ctnl_notifier_exp = {
1469         .notifier_call  = ctnetlink_expect_event,
1470 };
1471 #endif
1472
1473 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1474         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1475                                             .attr_count = CTA_MAX, },
1476         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1477                                             .attr_count = CTA_MAX, },
1478         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1479                                             .attr_count = CTA_MAX, },
1480         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1481                                             .attr_count = CTA_MAX, },
1482 };
1483
1484 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1485         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1486                                             .attr_count = CTA_EXPECT_MAX, },
1487         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1488                                             .attr_count = CTA_EXPECT_MAX, },
1489         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1490                                             .attr_count = CTA_EXPECT_MAX, },
1491 };
1492
1493 static struct nfnetlink_subsystem ctnl_subsys = {
1494         .name                           = "conntrack",
1495         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1496         .cb_count                       = IPCTNL_MSG_MAX,
1497         .cb                             = ctnl_cb,
1498 };
1499
1500 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1501         .name                           = "conntrack_expect",
1502         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1503         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1504         .cb                             = ctnl_exp_cb,
1505 };
1506
1507 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1508 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1509
1510 static int __init ctnetlink_init(void)
1511 {
1512         int ret;
1513
1514         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1515         ret = nfnetlink_subsys_register(&ctnl_subsys);
1516         if (ret < 0) {
1517                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1518                 goto err_out;
1519         }
1520
1521         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1522         if (ret < 0) {
1523                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1524                 goto err_unreg_subsys;
1525         }
1526
1527 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1528         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1529         if (ret < 0) {
1530                 printk("ctnetlink_init: cannot register notifier.\n");
1531                 goto err_unreg_exp_subsys;
1532         }
1533
1534         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1535         if (ret < 0) {
1536                 printk("ctnetlink_init: cannot expect register notifier.\n");
1537                 goto err_unreg_notifier;
1538         }
1539 #endif
1540
1541         return 0;
1542
1543 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1544 err_unreg_notifier:
1545         ip_conntrack_unregister_notifier(&ctnl_notifier);
1546 err_unreg_exp_subsys:
1547         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1548 #endif
1549 err_unreg_subsys:
1550         nfnetlink_subsys_unregister(&ctnl_subsys);
1551 err_out:
1552         return ret;
1553 }
1554
1555 static void __exit ctnetlink_exit(void)
1556 {
1557         printk("ctnetlink: unregistering from nfnetlink.\n");
1558
1559 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1560         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1561         ip_conntrack_unregister_notifier(&ctnl_notifier);
1562 #endif
1563
1564         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1565         nfnetlink_subsys_unregister(&ctnl_subsys);
1566         return;
1567 }
1568
1569 module_init(ctnetlink_init);
1570 module_exit(ctnetlink_exit);