SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv4/netfilter/ip_conntrack_netlink.c

   1 /* Connection tracking via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
   7  * (C) 2005-2006 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
 747         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 748                                   IPCTNL_MSG_CT_NEW, 1, ct);
 749         ip_conntrack_put(ct);
 750         if (err <= 0)
 751                 goto free;
 752
 753         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 754         if (err < 0)
 755                 goto out;
 756
 757         return 0;
 758
 759 free:
 760         kfree_skb(skb2);
 761 out:
 762         return err;
 763 }
 764
 765 static inline int
 766 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 767 {
 768         unsigned long d;
 769         unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
 770         d = ct->status ^ status;
 771
 772         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 773                 /* unchangeable */
 774                 return -EINVAL;
 775
 776         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 777                 /* SEEN_REPLY bit can only be set */
 778                 return -EINVAL;
 779
 780
 781         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 782                 /* ASSURED bit can only be set */
 783                 return -EINVAL;
 784
 785         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
 786 #ifndef CONFIG_IP_NF_NAT_NEEDED
 787                 return -EINVAL;
 788 #else
 789                 struct ip_nat_range range;
 790
 791                 if (cda[CTA_NAT_DST-1]) {
 792                         if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
 793                                                 &range) < 0)
 794                                 return -EINVAL;
 795                         if (ip_nat_initialized(ct,
 796                                                HOOK2MANIP(NF_IP_PRE_ROUTING)))
 797                                 return -EEXIST;
 798                         ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
 799                 }
 800                 if (cda[CTA_NAT_SRC-1]) {
 801                         if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
 802                                                 &range) < 0)
 803                                 return -EINVAL;
 804                         if (ip_nat_initialized(ct,
 805                                                HOOK2MANIP(NF_IP_POST_ROUTING)))
 806                                 return -EEXIST;
 807                         ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
 808                 }
 809 #endif
 810         }
 811
 812         /* Be careful here, modifying NAT bits can screw up things,
 813          * so don't let users modify them directly if they don't pass
 814          * ip_nat_range. */
 815         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 816         return 0;
 817 }
 818
 819
 820 static inline int
 821 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 822 {
 823         struct ip_conntrack_helper *helper;
 824         char *helpname;
 825         int err;
 826
 827         /* don't change helper of sibling connections */
 828         if (ct->master)
 829                 return -EINVAL;
 830
 831         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 832         if (err < 0)
 833                 return err;
 834
 835         helper = __ip_conntrack_helper_find_byname(helpname);
 836         if (!helper) {
 837                 if (!strcmp(helpname, ""))
 838                         helper = NULL;
 839                 else
 840                         return -EINVAL;
 841         }
 842
 843         if (ct->helper) {
 844                 if (!helper) {
 845                         /* we had a helper before ... */
 846                         ip_ct_remove_expectations(ct);
 847                         ct->helper = NULL;
 848                 } else {
 849                         /* need to zero data of old helper */
 850                         memset(&ct->help, 0, sizeof(ct->help));
 851                 }
 852         }
 853
 854         ct->helper = helper;
 855
 856         return 0;
 857 }
 858
 859 static inline int
 860 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 861 {
 862         u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 863
 864         if (!del_timer(&ct->timeout))
 865                 return -ETIME;
 866
 867         ct->timeout.expires = jiffies + timeout * HZ;
 868         add_timer(&ct->timeout);
 869
 870         return 0;
 871 }
 872
 873 static inline int
 874 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 875 {
 876         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 877         struct ip_conntrack_protocol *proto;
 878         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 879         int err = 0;
 880
 881         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 882
 883         proto = ip_conntrack_proto_find_get(npt);
 884
 885         if (proto->from_nfattr)
 886                 err = proto->from_nfattr(tb, ct);
 887         ip_conntrack_proto_put(proto);
 888
 889         return err;
 890 }
 891
 892 static int
 893 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 894 {
 895         int err;
 896
 897         if (cda[CTA_HELP-1]) {
 898                 err = ctnetlink_change_helper(ct, cda);
 899                 if (err < 0)
 900                         return err;
 901         }
 902
 903         if (cda[CTA_TIMEOUT-1]) {
 904                 err = ctnetlink_change_timeout(ct, cda);
 905                 if (err < 0)
 906                         return err;
 907         }
 908
 909         if (cda[CTA_STATUS-1]) {
 910                 err = ctnetlink_change_status(ct, cda);
 911                 if (err < 0)
 912                         return err;
 913         }
 914
 915         if (cda[CTA_PROTOINFO-1]) {
 916                 err = ctnetlink_change_protoinfo(ct, cda);
 917                 if (err < 0)
 918                         return err;
 919         }
 920
 921 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 922         if (cda[CTA_MARK-1])
 923                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 924 #endif
 925
 926         return 0;
 927 }
 928
 929 static int
 930 ctnetlink_create_conntrack(struct nfattr *cda[],
 931                            struct ip_conntrack_tuple *otuple,
 932                            struct ip_conntrack_tuple *rtuple)
 933 {
 934         struct ip_conntrack *ct;
 935         int err = -EINVAL;
 936
 937         ct = ip_conntrack_alloc(otuple, rtuple);
 938         if (ct == NULL || IS_ERR(ct))
 939                 return -ENOMEM;
 940
 941         if (!cda[CTA_TIMEOUT-1])
 942                 goto err;
 943         ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 944
 945         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
 946         ct->status |= IPS_CONFIRMED;
 947
 948         err = ctnetlink_change_status(ct, cda);
 949         if (err < 0)
 950                 goto err;
 951
 952         if (cda[CTA_PROTOINFO-1]) {
 953                 err = ctnetlink_change_protoinfo(ct, cda);
 954                 if (err < 0)
 955                         return err;
 956         }
 957
 958 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 959         if (cda[CTA_MARK-1])
 960                 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
 961 #endif
 962
 963         ct->helper = ip_conntrack_helper_find_get(rtuple);
 964
 965         add_timer(&ct->timeout);
 966         ip_conntrack_hash_insert(ct);
 967
 968         if (ct->helper)
 969                 ip_conntrack_helper_put(ct->helper);
 970
 971         return 0;
 972
 973 err:
 974         ip_conntrack_free(ct);
 975         return err;
 976 }
 977
 978 static int
 979 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
 980                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 981 {
 982         struct ip_conntrack_tuple otuple, rtuple;
 983         struct ip_conntrack_tuple_hash *h = NULL;
 984         int err = 0;
 985
 986         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 987                 return -EINVAL;
 988
 989         if (cda[CTA_TUPLE_ORIG-1]) {
 990                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
 991                 if (err < 0)
 992                         return err;
 993         }
 994
 995         if (cda[CTA_TUPLE_REPLY-1]) {
 996                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
 997                 if (err < 0)
 998                         return err;
 999         }
1000
1001         write_lock_bh(&ip_conntrack_lock);
1002         if (cda[CTA_TUPLE_ORIG-1])
1003                 h = __ip_conntrack_find(&otuple, NULL);
1004         else if (cda[CTA_TUPLE_REPLY-1])
1005                 h = __ip_conntrack_find(&rtuple, NULL);
1006
1007         if (h == NULL) {
1008                 write_unlock_bh(&ip_conntrack_lock);
1009                 err = -ENOENT;
1010                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1011                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1012                 return err;
1013         }
1014         /* implicit 'else' */
1015
1016         /* we only allow nat config for new conntracks */
1017         if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1018                 err = -EINVAL;
1019                 goto out_unlock;
1020         }
1021
1022         /* We manipulate the conntrack inside the global conntrack table lock,
1023          * so there's no need to increase the refcount */
1024         err = -EEXIST;
1025         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1026                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1027
1028 out_unlock:
1029         write_unlock_bh(&ip_conntrack_lock);
1030         return err;
1031 }
1032
1033 /***********************************************************************
1034  * EXPECT
1035  ***********************************************************************/
1036
1037 static inline int
1038 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1039                          const struct ip_conntrack_tuple *tuple,
1040                          enum ctattr_expect type)
1041 {
1042         struct nfattr *nest_parms = NFA_NEST(skb, type);
1043
1044         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1045                 goto nfattr_failure;
1046
1047         NFA_NEST_END(skb, nest_parms);
1048
1049         return 0;
1050
1051 nfattr_failure:
1052         return -1;
1053 }
1054
1055 static inline int
1056 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1057                         const struct ip_conntrack_tuple *tuple,
1058                         const struct ip_conntrack_tuple *mask)
1059 {
1060         int ret;
1061         struct ip_conntrack_protocol *proto;
1062         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1063
1064         ret = ctnetlink_dump_tuples_ip(skb, mask);
1065         if (unlikely(ret < 0))
1066                 goto nfattr_failure;
1067
1068         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1069         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1070         ip_conntrack_proto_put(proto);
1071         if (unlikely(ret < 0))
1072                 goto nfattr_failure;
1073
1074         NFA_NEST_END(skb, nest_parms);
1075
1076         return 0;
1077
1078 nfattr_failure:
1079         return -1;
1080 }
1081
1082 static inline int
1083 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1084                           const struct ip_conntrack_expect *exp)
1085 {
1086         struct ip_conntrack *master = exp->master;
1087         __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1088         __be32 id = htonl(exp->id);
1089
1090         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1091                 goto nfattr_failure;
1092         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1093                 goto nfattr_failure;
1094         if (ctnetlink_exp_dump_tuple(skb,
1095                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1096                                  CTA_EXPECT_MASTER) < 0)
1097                 goto nfattr_failure;
1098
1099         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1100         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1101
1102         return 0;
1103
1104 nfattr_failure:
1105         return -1;
1106 }
1107
1108 static int
1109 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1110                     int event,
1111                     int nowait,
1112                     const struct ip_conntrack_expect *exp)
1113 {
1114         struct nlmsghdr *nlh;
1115         struct nfgenmsg *nfmsg;
1116         unsigned char *b;
1117
1118         b = skb->tail;
1119
1120         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1121         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1122         nfmsg  = NLMSG_DATA(nlh);
1123
1124         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1125         nfmsg->nfgen_family = AF_INET;
1126         nfmsg->version      = NFNETLINK_V0;
1127         nfmsg->res_id       = 0;
1128
1129         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1130                 goto nfattr_failure;
1131
1132         nlh->nlmsg_len = skb->tail - b;
1133         return skb->len;
1134
1135 nlmsg_failure:
1136 nfattr_failure:
1137         skb_trim(skb, b - skb->data);
1138         return -1;
1139 }
1140
1141 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1142 static int ctnetlink_expect_event(struct notifier_block *this,
1143                                   unsigned long events, void *ptr)
1144 {
1145         struct nlmsghdr *nlh;
1146         struct nfgenmsg *nfmsg;
1147         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1148         struct sk_buff *skb;
1149         unsigned int type;
1150         unsigned char *b;
1151         int flags = 0;
1152
1153         if (events & IPEXP_NEW) {
1154                 type = IPCTNL_MSG_EXP_NEW;
1155                 flags = NLM_F_CREATE|NLM_F_EXCL;
1156         } else
1157                 return NOTIFY_DONE;
1158
1159         if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1160                 return NOTIFY_DONE;
1161
1162         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1163         if (!skb)
1164                 return NOTIFY_DONE;
1165
1166         b = skb->tail;
1167
1168         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1169         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1170         nfmsg = NLMSG_DATA(nlh);
1171
1172         nlh->nlmsg_flags    = flags;
1173         nfmsg->nfgen_family = AF_INET;
1174         nfmsg->version      = NFNETLINK_V0;
1175         nfmsg->res_id       = 0;
1176
1177         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1178                 goto nfattr_failure;
1179
1180         nlh->nlmsg_len = skb->tail - b;
1181         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1182         return NOTIFY_DONE;
1183
1184 nlmsg_failure:
1185 nfattr_failure:
1186         kfree_skb(skb);
1187         return NOTIFY_DONE;
1188 }
1189 #endif
1190
1191 static int
1192 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1193 {
1194         struct ip_conntrack_expect *exp = NULL;
1195         struct list_head *i;
1196         u_int32_t *id = (u_int32_t *) &cb->args[0];
1197
1198         read_lock_bh(&ip_conntrack_lock);
1199         list_for_each_prev(i, &ip_conntrack_expect_list) {
1200                 exp = (struct ip_conntrack_expect *) i;
1201                 if (exp->id <= *id)
1202                         continue;
1203                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1204                                             cb->nlh->nlmsg_seq,
1205                                             IPCTNL_MSG_EXP_NEW,
1206                                             1, exp) < 0)
1207                         goto out;
1208                 *id = exp->id;
1209         }
1210 out:
1211         read_unlock_bh(&ip_conntrack_lock);
1212
1213         return skb->len;
1214 }
1215
1216 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1217         [CTA_EXPECT_TIMEOUT-1]          = sizeof(__be32),
1218         [CTA_EXPECT_ID-1]               = sizeof(__be32)
1219 };
1220
1221 static int
1222 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1223                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1224 {
1225         struct ip_conntrack_tuple tuple;
1226         struct ip_conntrack_expect *exp;
1227         struct sk_buff *skb2;
1228         int err = 0;
1229
1230         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1231                 return -EINVAL;
1232
1233         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1234                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1235                 u32 rlen;
1236
1237                 if (msg->nfgen_family != AF_INET)
1238                         return -EAFNOSUPPORT;
1239
1240                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1241                                                 ctnetlink_exp_dump_table,
1242                                                 ctnetlink_done)) != 0)
1243                         return -EINVAL;
1244                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1245                 if (rlen > skb->len)
1246                         rlen = skb->len;
1247                 skb_pull(skb, rlen);
1248                 return 0;
1249         }
1250
1251         if (cda[CTA_EXPECT_MASTER-1])
1252                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1253         else
1254                 return -EINVAL;
1255
1256         if (err < 0)
1257                 return err;
1258
1259         exp = ip_conntrack_expect_find_get(&tuple);
1260         if (!exp)
1261                 return -ENOENT;
1262
1263         if (cda[CTA_EXPECT_ID-1]) {
1264                 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1265                 if (exp->id != ntohl(id)) {
1266                         ip_conntrack_expect_put(exp);
1267                         return -ENOENT;
1268                 }
1269         }
1270
1271         err = -ENOMEM;
1272         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1273         if (!skb2)
1274                 goto out;
1275
1276         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1277                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1278                                       1, exp);
1279         if (err <= 0)
1280                 goto free;
1281
1282         ip_conntrack_expect_put(exp);
1283
1284         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1285
1286 free:
1287         kfree_skb(skb2);
1288 out:
1289         ip_conntrack_expect_put(exp);
1290         return err;
1291 }
1292
1293 static int
1294 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1295                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1296 {
1297         struct ip_conntrack_expect *exp, *tmp;
1298         struct ip_conntrack_tuple tuple;
1299         struct ip_conntrack_helper *h;
1300         int err;
1301
1302         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1303                 return -EINVAL;
1304
1305         if (cda[CTA_EXPECT_TUPLE-1]) {
1306                 /* delete a single expect by tuple */
1307                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1308                 if (err < 0)
1309                         return err;
1310
1311                 /* bump usage count to 2 */
1312                 exp = ip_conntrack_expect_find_get(&tuple);
1313                 if (!exp)
1314                         return -ENOENT;
1315
1316                 if (cda[CTA_EXPECT_ID-1]) {
1317                         __be32 id =
1318                                 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1319                         if (exp->id != ntohl(id)) {
1320                                 ip_conntrack_expect_put(exp);
1321                                 return -ENOENT;
1322                         }
1323                 }
1324
1325                 /* after list removal, usage count == 1 */
1326                 ip_conntrack_unexpect_related(exp);
1327                 /* have to put what we 'get' above.
1328                  * after this line usage count == 0 */
1329                 ip_conntrack_expect_put(exp);
1330         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1331                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1332
1333                 /* delete all expectations for this helper */
1334                 write_lock_bh(&ip_conntrack_lock);
1335                 h = __ip_conntrack_helper_find_byname(name);
1336                 if (!h) {
1337                         write_unlock_bh(&ip_conntrack_lock);
1338                         return -EINVAL;
1339                 }
1340                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1341                                          list) {
1342                         if (exp->master->helper == h
1343                             && del_timer(&exp->timeout)) {
1344                                 ip_ct_unlink_expect(exp);
1345                                 ip_conntrack_expect_put(exp);
1346                         }
1347                 }
1348                 write_unlock_bh(&ip_conntrack_lock);
1349         } else {
1350                 /* This basically means we have to flush everything*/
1351                 write_lock_bh(&ip_conntrack_lock);
1352                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1353                                          list) {
1354                         if (del_timer(&exp->timeout)) {
1355                                 ip_ct_unlink_expect(exp);
1356                                 ip_conntrack_expect_put(exp);
1357                         }
1358                 }
1359                 write_unlock_bh(&ip_conntrack_lock);
1360         }
1361
1362         return 0;
1363 }
1364 static int
1365 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1366 {
1367         return -EOPNOTSUPP;
1368 }
1369
1370 static int
1371 ctnetlink_create_expect(struct nfattr *cda[])
1372 {
1373         struct ip_conntrack_tuple tuple, mask, master_tuple;
1374         struct ip_conntrack_tuple_hash *h = NULL;
1375         struct ip_conntrack_expect *exp;
1376         struct ip_conntrack *ct;
1377         int err = 0;
1378
1379         /* caller guarantees that those three CTA_EXPECT_* exist */
1380         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1381         if (err < 0)
1382                 return err;
1383         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1384         if (err < 0)
1385                 return err;
1386         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1387         if (err < 0)
1388                 return err;
1389
1390         /* Look for master conntrack of this expectation */
1391         h = ip_conntrack_find_get(&master_tuple, NULL);
1392         if (!h)
1393                 return -ENOENT;
1394         ct = tuplehash_to_ctrack(h);
1395
1396         if (!ct->helper) {
1397                 /* such conntrack hasn't got any helper, abort */
1398                 err = -EINVAL;
1399                 goto out;
1400         }
1401
1402         exp = ip_conntrack_expect_alloc(ct);
1403         if (!exp) {
1404                 err = -ENOMEM;
1405                 goto out;
1406         }
1407
1408         exp->expectfn = NULL;
1409         exp->flags = 0;
1410         exp->master = ct;
1411         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1412         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1413
1414         err = ip_conntrack_expect_related(exp);
1415         ip_conntrack_expect_put(exp);
1416
1417 out:
1418         ip_conntrack_put(tuplehash_to_ctrack(h));
1419         return err;
1420 }
1421
1422 static int
1423 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1424                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1425 {
1426         struct ip_conntrack_tuple tuple;
1427         struct ip_conntrack_expect *exp;
1428         int err = 0;
1429
1430         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1431                 return -EINVAL;
1432
1433         if (!cda[CTA_EXPECT_TUPLE-1]
1434             || !cda[CTA_EXPECT_MASK-1]
1435             || !cda[CTA_EXPECT_MASTER-1])
1436                 return -EINVAL;
1437
1438         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1439         if (err < 0)
1440                 return err;
1441
1442         write_lock_bh(&ip_conntrack_lock);
1443         exp = __ip_conntrack_expect_find(&tuple);
1444
1445         if (!exp) {
1446                 write_unlock_bh(&ip_conntrack_lock);
1447                 err = -ENOENT;
1448                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1449                         err = ctnetlink_create_expect(cda);
1450                 return err;
1451         }
1452
1453         err = -EEXIST;
1454         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1455                 err = ctnetlink_change_expect(exp, cda);
1456         write_unlock_bh(&ip_conntrack_lock);
1457
1458         return err;
1459 }
1460
1461 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1462 static struct notifier_block ctnl_notifier = {
1463         .notifier_call  = ctnetlink_conntrack_event,
1464 };
1465
1466 static struct notifier_block ctnl_notifier_exp = {
1467         .notifier_call  = ctnetlink_expect_event,
1468 };
1469 #endif
1470
1471 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1472         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1473                                             .attr_count = CTA_MAX, },
1474         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1475                                             .attr_count = CTA_MAX, },
1476         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1477                                             .attr_count = CTA_MAX, },
1478         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1479                                             .attr_count = CTA_MAX, },
1480 };
1481
1482 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1483         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1484                                             .attr_count = CTA_EXPECT_MAX, },
1485         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1486                                             .attr_count = CTA_EXPECT_MAX, },
1487         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1488                                             .attr_count = CTA_EXPECT_MAX, },
1489 };
1490
1491 static struct nfnetlink_subsystem ctnl_subsys = {
1492         .name                           = "conntrack",
1493         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1494         .cb_count                       = IPCTNL_MSG_MAX,
1495         .cb                             = ctnl_cb,
1496 };
1497
1498 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1499         .name                           = "conntrack_expect",
1500         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1501         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1502         .cb                             = ctnl_exp_cb,
1503 };
1504
1505 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1506 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1507
1508 static int __init ctnetlink_init(void)
1509 {
1510         int ret;
1511
1512         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1513         ret = nfnetlink_subsys_register(&ctnl_subsys);
1514         if (ret < 0) {
1515                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1516                 goto err_out;
1517         }
1518
1519         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1520         if (ret < 0) {
1521                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1522                 goto err_unreg_subsys;
1523         }
1524
1525 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1526         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1527         if (ret < 0) {
1528                 printk("ctnetlink_init: cannot register notifier.\n");
1529                 goto err_unreg_exp_subsys;
1530         }
1531
1532         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1533         if (ret < 0) {
1534                 printk("ctnetlink_init: cannot expect register notifier.\n");
1535                 goto err_unreg_notifier;
1536         }
1537 #endif
1538
1539         return 0;
1540
1541 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1542 err_unreg_notifier:
1543         ip_conntrack_unregister_notifier(&ctnl_notifier);
1544 err_unreg_exp_subsys:
1545         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1546 #endif
1547 err_unreg_subsys:
1548         nfnetlink_subsys_unregister(&ctnl_subsys);
1549 err_out:
1550         return ret;
1551 }
1552
1553 static void __exit ctnetlink_exit(void)
1554 {
1555         printk("ctnetlink: unregistering from nfnetlink.\n");
1556
1557 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1558         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1559         ip_conntrack_unregister_notifier(&ctnl_notifier);
1560 #endif
1561
1562         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1563         nfnetlink_subsys_unregister(&ctnl_subsys);
1564         return;
1565 }
1566
1567 module_init(ctnetlink_init);
1568 module_exit(ctnetlink_exit);