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 #if 0
  48 #define DEBUGP printk
  49 #else
  50 #define DEBUGP(format, args...)
  51 #endif
  52
  53
  54 static inline int
  55 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
  56                             const struct ip_conntrack_tuple *tuple,
  57                             struct ip_conntrack_protocol *proto)
  58 {
  59         int ret = 0;
  60         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  61
  62         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  63
  64         if (likely(proto->tuple_to_nfattr))
  65                 ret = proto->tuple_to_nfattr(skb, tuple);
  66
  67         NFA_NEST_END(skb, nest_parms);
  68
  69         return ret;
  70
  71 nfattr_failure:
  72         return -1;
  73 }
  74
  75 static inline int
  76 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
  77                          const struct ip_conntrack_tuple *tuple)
  78 {
  79         struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  80
  81         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  82         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  83
  84         NFA_NEST_END(skb, nest_parms);
  85
  86         return 0;
  87
  88 nfattr_failure:
  89         return -1;
  90 }
  91
  92 static inline int
  93 ctnetlink_dump_tuples(struct sk_buff *skb,
  94                       const struct ip_conntrack_tuple *tuple)
  95 {
  96         int ret;
  97         struct ip_conntrack_protocol *proto;
  98
  99         ret = ctnetlink_dump_tuples_ip(skb, tuple);
 100         if (unlikely(ret < 0))
 101                 return ret;
 102
 103         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 104         ret = ctnetlink_dump_tuples_proto(skb, tuple, proto);
 105         ip_conntrack_proto_put(proto);
 106
 107         return ret;
 108 }
 109
 110 static inline int
 111 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 112 {
 113         u_int32_t status = htonl((u_int32_t) ct->status);
 114         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 115         return 0;
 116
 117 nfattr_failure:
 118         return -1;
 119 }
 120
 121 static inline int
 122 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 123 {
 124         long timeout_l = ct->timeout.expires - jiffies;
 125         u_int32_t timeout;
 126
 127         if (timeout_l < 0)
 128                 timeout = 0;
 129         else
 130                 timeout = htonl(timeout_l / HZ);
 131
 132         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 133         return 0;
 134
 135 nfattr_failure:
 136         return -1;
 137 }
 138
 139 static inline int
 140 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 141 {
 142         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 143
 144         struct nfattr *nest_proto;
 145         int ret;
 146
 147         if (!proto->to_nfattr) {
 148                 ip_conntrack_proto_put(proto);
 149                 return 0;
 150         }
 151
 152         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 153
 154         ret = proto->to_nfattr(skb, nest_proto, ct);
 155
 156         ip_conntrack_proto_put(proto);
 157
 158         NFA_NEST_END(skb, nest_proto);
 159
 160         return ret;
 161
 162 nfattr_failure:
 163         return -1;
 164 }
 165
 166 static inline int
 167 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 168 {
 169         struct nfattr *nest_helper;
 170
 171         if (!ct->helper)
 172                 return 0;
 173
 174         nest_helper = NFA_NEST(skb, CTA_HELP);
 175         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 176
 177         if (ct->helper->to_nfattr)
 178                 ct->helper->to_nfattr(skb, ct);
 179
 180         NFA_NEST_END(skb, nest_helper);
 181
 182         return 0;
 183
 184 nfattr_failure:
 185         return -1;
 186 }
 187
 188 #ifdef CONFIG_IP_NF_CT_ACCT
 189 static inline int
 190 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 191                         enum ip_conntrack_dir dir)
 192 {
 193         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 194         struct nfattr *nest_count = NFA_NEST(skb, type);
 195         u_int32_t tmp;
 196
 197         tmp = htonl(ct->counters[dir].packets);
 198         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 199
 200         tmp = htonl(ct->counters[dir].bytes);
 201         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 202
 203         NFA_NEST_END(skb, nest_count);
 204
 205         return 0;
 206
 207 nfattr_failure:
 208         return -1;
 209 }
 210 #else
 211 #define ctnetlink_dump_counters(a, b, c) (0)
 212 #endif
 213
 214 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 215 static inline int
 216 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 217 {
 218         u_int32_t mark = htonl(ct->mark);
 219
 220         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 221         return 0;
 222
 223 nfattr_failure:
 224         return -1;
 225 }
 226 #else
 227 #define ctnetlink_dump_mark(a, b) (0)
 228 #endif
 229
 230 static inline int
 231 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 232 {
 233         u_int32_t id = htonl(ct->id);
 234         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 235         return 0;
 236
 237 nfattr_failure:
 238         return -1;
 239 }
 240
 241 static inline int
 242 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 243 {
 244         u_int32_t use = htonl(atomic_read(&ct->ct_general.use));
 245
 246         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 247         return 0;
 248
 249 nfattr_failure:
 250         return -1;
 251 }
 252
 253 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 254
 255 static int
 256 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 257                     int event, int nowait,
 258                     const struct ip_conntrack *ct)
 259 {
 260         struct nlmsghdr *nlh;
 261         struct nfgenmsg *nfmsg;
 262         struct nfattr *nest_parms;
 263         unsigned char *b;
 264
 265         b = skb->tail;
 266
 267         event |= NFNL_SUBSYS_CTNETLINK << 8;
 268         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 269         nfmsg  = NLMSG_DATA(nlh);
 270
 271         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 272         nfmsg->nfgen_family = AF_INET;
 273         nfmsg->version      = NFNETLINK_V0;
 274         nfmsg->res_id       = 0;
 275
 276         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 277         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 278                 goto nfattr_failure;
 279         NFA_NEST_END(skb, nest_parms);
 280
 281         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 282         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 283                 goto nfattr_failure;
 284         NFA_NEST_END(skb, nest_parms);
 285
 286         if (ctnetlink_dump_status(skb, ct) < 0 ||
 287             ctnetlink_dump_timeout(skb, ct) < 0 ||
 288             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 289             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 290             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 291             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 292             ctnetlink_dump_mark(skb, ct) < 0 ||
 293             ctnetlink_dump_id(skb, ct) < 0 ||
 294             ctnetlink_dump_use(skb, ct) < 0)
 295                 goto nfattr_failure;
 296
 297         nlh->nlmsg_len = skb->tail - b;
 298         return skb->len;
 299
 300 nlmsg_failure:
 301 nfattr_failure:
 302         skb_trim(skb, b - skb->data);
 303         return -1;
 304 }
 305
 306 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 307 static int ctnetlink_conntrack_event(struct notifier_block *this,
 308                                      unsigned long events, void *ptr)
 309 {
 310         struct nlmsghdr *nlh;
 311         struct nfgenmsg *nfmsg;
 312         struct nfattr *nest_parms;
 313         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 314         struct sk_buff *skb;
 315         unsigned int type;
 316         unsigned char *b;
 317         unsigned int flags = 0, group;
 318
 319         /* ignore our fake conntrack entry */
 320         if (ct == &ip_conntrack_untracked)
 321                 return NOTIFY_DONE;
 322
 323         if (events & IPCT_DESTROY) {
 324                 type = IPCTNL_MSG_CT_DELETE;
 325                 group = NFNLGRP_CONNTRACK_DESTROY;
 326         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 327                 type = IPCTNL_MSG_CT_NEW;
 328                 flags = NLM_F_CREATE|NLM_F_EXCL;
 329                 /* dump everything */
 330                 events = ~0UL;
 331                 group = NFNLGRP_CONNTRACK_NEW;
 332         } else if (events & (IPCT_STATUS |
 333                       IPCT_PROTOINFO |
 334                       IPCT_HELPER |
 335                       IPCT_HELPINFO |
 336                       IPCT_NATINFO)) {
 337                 type = IPCTNL_MSG_CT_NEW;
 338                 group = NFNLGRP_CONNTRACK_UPDATE;
 339         } else
 340                 return NOTIFY_DONE;
 341
 342         if (!nfnetlink_has_listeners(group))
 343                 return NOTIFY_DONE;
 344
 345         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 346         if (!skb)
 347                 return NOTIFY_DONE;
 348
 349         b = skb->tail;
 350
 351         type |= NFNL_SUBSYS_CTNETLINK << 8;
 352         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 353         nfmsg = NLMSG_DATA(nlh);
 354
 355         nlh->nlmsg_flags    = flags;
 356         nfmsg->nfgen_family = AF_INET;
 357         nfmsg->version  = NFNETLINK_V0;
 358         nfmsg->res_id   = 0;
 359
 360         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 361         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 362                 goto nfattr_failure;
 363         NFA_NEST_END(skb, nest_parms);
 364
 365         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 366         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 367                 goto nfattr_failure;
 368         NFA_NEST_END(skb, nest_parms);
 369
 370         /* NAT stuff is now a status flag */
 371         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 372             && ctnetlink_dump_status(skb, ct) < 0)
 373                 goto nfattr_failure;
 374         if (events & IPCT_REFRESH
 375             && ctnetlink_dump_timeout(skb, ct) < 0)
 376                 goto nfattr_failure;
 377         if (events & IPCT_PROTOINFO
 378             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 379                 goto nfattr_failure;
 380         if (events & IPCT_HELPINFO
 381             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 382                 goto nfattr_failure;
 383
 384         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 385             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 386                 goto nfattr_failure;
 387
 388         nlh->nlmsg_len = skb->tail - b;
 389         nfnetlink_send(skb, 0, group, 0);
 390         return NOTIFY_DONE;
 391
 392 nlmsg_failure:
 393 nfattr_failure:
 394         kfree_skb(skb);
 395         return NOTIFY_DONE;
 396 }
 397 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 398
 399 static int ctnetlink_done(struct netlink_callback *cb)
 400 {
 401         DEBUGP("entered %s\n", __FUNCTION__);
 402         return 0;
 403 }
 404
 405 static int
 406 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 407 {
 408         struct ip_conntrack *ct = NULL;
 409         struct ip_conntrack_tuple_hash *h;
 410         struct list_head *i;
 411         u_int32_t *id = (u_int32_t *) &cb->args[1];
 412
 413         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 414                         cb->args[0], *id);
 415
 416         read_lock_bh(&ip_conntrack_lock);
 417         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 418                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 419                         h = (struct ip_conntrack_tuple_hash *) i;
 420                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 421                                 continue;
 422                         ct = tuplehash_to_ctrack(h);
 423                         if (ct->id <= *id)
 424                                 continue;
 425                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 426                                                 cb->nlh->nlmsg_seq,
 427                                                 IPCTNL_MSG_CT_NEW,
 428                                                 1, ct) < 0)
 429                                 goto out;
 430                         *id = ct->id;
 431                 }
 432         }
 433 out:
 434         read_unlock_bh(&ip_conntrack_lock);
 435
 436         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 437
 438         return skb->len;
 439 }
 440
 441 #ifdef CONFIG_IP_NF_CT_ACCT
 442 static int
 443 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 444 {
 445         struct ip_conntrack *ct = NULL;
 446         struct ip_conntrack_tuple_hash *h;
 447         struct list_head *i;
 448         u_int32_t *id = (u_int32_t *) &cb->args[1];
 449
 450         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 451                         cb->args[0], *id);
 452
 453         write_lock_bh(&ip_conntrack_lock);
 454         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 455                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 456                         h = (struct ip_conntrack_tuple_hash *) i;
 457                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 458                                 continue;
 459                         ct = tuplehash_to_ctrack(h);
 460                         if (ct->id <= *id)
 461                                 continue;
 462                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 463                                                 cb->nlh->nlmsg_seq,
 464                                                 IPCTNL_MSG_CT_NEW,
 465                                                 1, ct) < 0)
 466                                 goto out;
 467                         *id = ct->id;
 468
 469                         memset(&ct->counters, 0, sizeof(ct->counters));
 470                 }
 471         }
 472 out:
 473         write_unlock_bh(&ip_conntrack_lock);
 474
 475         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 476
 477         return skb->len;
 478 }
 479 #endif
 480
 481 static const size_t cta_min_ip[CTA_IP_MAX] = {
 482         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 483         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 484 };
 485
 486 static inline int
 487 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 488 {
 489         struct nfattr *tb[CTA_IP_MAX];
 490
 491         DEBUGP("entered %s\n", __FUNCTION__);
 492
 493         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 494
 495         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 496                 return -EINVAL;
 497
 498         if (!tb[CTA_IP_V4_SRC-1])
 499                 return -EINVAL;
 500         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 501
 502         if (!tb[CTA_IP_V4_DST-1])
 503                 return -EINVAL;
 504         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 505
 506         DEBUGP("leaving\n");
 507
 508         return 0;
 509 }
 510
 511 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 512         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 513         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 514         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 515         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 516         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 517         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 518 };
 519
 520 static inline int
 521 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 522                             struct ip_conntrack_tuple *tuple)
 523 {
 524         struct nfattr *tb[CTA_PROTO_MAX];
 525         struct ip_conntrack_protocol *proto;
 526         int ret = 0;
 527
 528         DEBUGP("entered %s\n", __FUNCTION__);
 529
 530         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 531
 532         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 533                 return -EINVAL;
 534
 535         if (!tb[CTA_PROTO_NUM-1])
 536                 return -EINVAL;
 537         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 538
 539         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 540
 541         if (likely(proto->nfattr_to_tuple))
 542                 ret = proto->nfattr_to_tuple(tb, tuple);
 543
 544         ip_conntrack_proto_put(proto);
 545
 546         return ret;
 547 }
 548
 549 static inline int
 550 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 551                       enum ctattr_tuple type)
 552 {
 553         struct nfattr *tb[CTA_TUPLE_MAX];
 554         int err;
 555
 556         DEBUGP("entered %s\n", __FUNCTION__);
 557
 558         memset(tuple, 0, sizeof(*tuple));
 559
 560         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 561
 562         if (!tb[CTA_TUPLE_IP-1])
 563                 return -EINVAL;
 564
 565         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 566         if (err < 0)
 567                 return err;
 568
 569         if (!tb[CTA_TUPLE_PROTO-1])
 570                 return -EINVAL;
 571
 572         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 573         if (err < 0)
 574                 return err;
 575
 576         /* orig and expect tuples get DIR_ORIGINAL */
 577         if (type == CTA_TUPLE_REPLY)
 578                 tuple->dst.dir = IP_CT_DIR_REPLY;
 579         else
 580                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 581
 582         DUMP_TUPLE(tuple);
 583
 584         DEBUGP("leaving\n");
 585
 586         return 0;
 587 }
 588
 589 #ifdef CONFIG_IP_NF_NAT_NEEDED
 590 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 591         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 592         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 593 };
 594
 595 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 596                                      const struct ip_conntrack *ct,
 597                                      struct ip_nat_range *range)
 598 {
 599         struct nfattr *tb[CTA_PROTONAT_MAX];
 600         struct ip_nat_protocol *npt;
 601
 602         DEBUGP("entered %s\n", __FUNCTION__);
 603
 604         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 605
 606         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 607                 return -EINVAL;
 608
 609         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 610
 611         if (!npt->nfattr_to_range) {
 612                 ip_nat_proto_put(npt);
 613                 return 0;
 614         }
 615
 616         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 617         if (npt->nfattr_to_range(tb, range) > 0)
 618                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 619
 620         ip_nat_proto_put(npt);
 621
 622         DEBUGP("leaving\n");
 623         return 0;
 624 }
 625
 626 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 627         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 628         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 629 };
 630
 631 static inline int
 632 ctnetlink_parse_nat(struct nfattr *cda[],
 633                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 634 {
 635         struct nfattr *tb[CTA_NAT_MAX];
 636         int err;
 637
 638         DEBUGP("entered %s\n", __FUNCTION__);
 639
 640         memset(range, 0, sizeof(*range));
 641
 642         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 643
 644         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 645                 return -EINVAL;
 646
 647         if (tb[CTA_NAT_MINIP-1])
 648                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 649
 650         if (!tb[CTA_NAT_MAXIP-1])
 651                 range->max_ip = range->min_ip;
 652         else
 653                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 654
 655         if (range->min_ip)
 656                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 657
 658         if (!tb[CTA_NAT_PROTO-1])
 659                 return 0;
 660
 661         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 662         if (err < 0)
 663                 return err;
 664
 665         DEBUGP("leaving\n");
 666         return 0;
 667 }
 668 #endif
 669
 670 static inline int
 671 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 672 {
 673         struct nfattr *tb[CTA_HELP_MAX];
 674
 675         DEBUGP("entered %s\n", __FUNCTION__);
 676
 677         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 678
 679         if (!tb[CTA_HELP_NAME-1])
 680                 return -EINVAL;
 681
 682         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 683
 684         return 0;
 685 }
 686
 687 static const size_t cta_min[CTA_MAX] = {
 688         [CTA_STATUS-1]          = sizeof(u_int32_t),
 689         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 690         [CTA_MARK-1]            = sizeof(u_int32_t),
 691         [CTA_USE-1]             = sizeof(u_int32_t),
 692         [CTA_ID-1]              = sizeof(u_int32_t)
 693 };
 694
 695 static int
 696 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 697                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 698 {
 699         struct ip_conntrack_tuple_hash *h;
 700         struct ip_conntrack_tuple tuple;
 701         struct ip_conntrack *ct;
 702         int err = 0;
 703
 704         DEBUGP("entered %s\n", __FUNCTION__);
 705
 706         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 707                 return -EINVAL;
 708
 709         if (cda[CTA_TUPLE_ORIG-1])
 710                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 711         else if (cda[CTA_TUPLE_REPLY-1])
 712                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 713         else {
 714                 /* Flush the whole table */
 715                 ip_conntrack_flush();
 716                 return 0;
 717         }
 718
 719         if (err < 0)
 720                 return err;
 721
 722         h = ip_conntrack_find_get(&tuple, NULL);
 723         if (!h) {
 724                 DEBUGP("tuple not found in conntrack hash\n");
 725                 return -ENOENT;
 726         }
 727
 728         ct = tuplehash_to_ctrack(h);
 729
 730         if (cda[CTA_ID-1]) {
 731                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 732                 if (ct->id != id) {
 733                         ip_conntrack_put(ct);
 734                         return -ENOENT;
 735                 }
 736         }
 737         if (del_timer(&ct->timeout))
 738                 ct->timeout.function((unsigned long)ct);
 739
 740         ip_conntrack_put(ct);
 741         DEBUGP("leaving\n");
 742
 743         return 0;
 744 }
 745
 746 static int
 747 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 748                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 749 {
 750         struct ip_conntrack_tuple_hash *h;
 751         struct ip_conntrack_tuple tuple;
 752         struct ip_conntrack *ct;
 753         struct sk_buff *skb2 = NULL;
 754         int err = 0;
 755
 756         DEBUGP("entered %s\n", __FUNCTION__);
 757
 758         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 759                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 760                 u32 rlen;
 761
 762                 if (msg->nfgen_family != AF_INET)
 763                         return -EAFNOSUPPORT;
 764
 765                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 766                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 767 #ifdef CONFIG_IP_NF_CT_ACCT
 768                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 769                                                 ctnetlink_dump_table_w,
 770                                                 ctnetlink_done)) != 0)
 771                                 return -EINVAL;
 772 #else
 773                         return -ENOTSUPP;
 774 #endif
 775                 } else {
 776                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 777                                                         ctnetlink_dump_table,
 778                                                         ctnetlink_done)) != 0)
 779                         return -EINVAL;
 780                 }
 781
 782                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 783                 if (rlen > skb->len)
 784                         rlen = skb->len;
 785                 skb_pull(skb, rlen);
 786                 return 0;
 787         }
 788
 789         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 790                 return -EINVAL;
 791
 792         if (cda[CTA_TUPLE_ORIG-1])
 793                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 794         else if (cda[CTA_TUPLE_REPLY-1])
 795                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 796         else
 797                 return -EINVAL;
 798
 799         if (err < 0)
 800                 return err;
 801
 802         h = ip_conntrack_find_get(&tuple, NULL);
 803         if (!h) {
 804                 DEBUGP("tuple not found in conntrack hash");
 805                 return -ENOENT;
 806         }
 807         DEBUGP("tuple found\n");
 808         ct = tuplehash_to_ctrack(h);
 809
 810         err = -ENOMEM;
 811         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 812         if (!skb2) {
 813                 ip_conntrack_put(ct);
 814                 return -ENOMEM;
 815         }
 816         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 817
 818         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 819                                   IPCTNL_MSG_CT_NEW, 1, ct);
 820         ip_conntrack_put(ct);
 821         if (err <= 0)
 822                 goto free;
 823
 824         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 825         if (err < 0)
 826                 goto out;
 827
 828         DEBUGP("leaving\n");
 829         return 0;
 830
 831 free:
 832         kfree_skb(skb2);
 833 out:
 834         return err;
 835 }
 836
 837 static inline int
 838 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 839 {
 840         unsigned long d;
 841         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 842         d = ct->status ^ status;
 843
 844         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 845                 /* unchangeable */
 846                 return -EINVAL;
 847
 848         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 849                 /* SEEN_REPLY bit can only be set */
 850                 return -EINVAL;
 851
 852
 853         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 854                 /* ASSURED bit can only be set */
 855                 return -EINVAL;
 856
 857         if (cda[CTA_NAT-1]) {
 858 #ifndef CONFIG_IP_NF_NAT_NEEDED
 859                 return -EINVAL;
 860 #else
 861                 unsigned int hooknum;
 862                 struct ip_nat_range range;
 863
 864                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 865                         return -EINVAL;
 866
 867                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 868                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 869                        htons(range.min.all), htons(range.max.all));
 870
 871                 /* This is tricky but it works. ip_nat_setup_info needs the
 872                  * hook number as parameter, so let's do the correct
 873                  * conversion and run away */
 874                 if (status & IPS_SRC_NAT_DONE)
 875                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 876                 else if (status & IPS_DST_NAT_DONE)
 877                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 878                 else
 879                         return -EINVAL; /* Missing NAT flags */
 880
 881                 DEBUGP("NAT status: %lu\n",
 882                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 883
 884                 if (ip_nat_initialized(ct, HOOK2MANIP(hooknum)))
 885                         return -EEXIST;
 886                 ip_nat_setup_info(ct, &range, hooknum);
 887
 888                 DEBUGP("NAT status after setup_info: %lu\n",
 889                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 890 #endif
 891         }
 892
 893         /* Be careful here, modifying NAT bits can screw up things,
 894          * so don't let users modify them directly if they don't pass
 895          * ip_nat_range. */
 896         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 897         return 0;
 898 }
 899
 900
 901 static inline int
 902 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 903 {
 904         struct ip_conntrack_helper *helper;
 905         char *helpname;
 906         int err;
 907
 908         DEBUGP("entered %s\n", __FUNCTION__);
 909
 910         /* don't change helper of sibling connections */
 911         if (ct->master)
 912                 return -EINVAL;
 913
 914         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 915         if (err < 0)
 916                 return err;
 917
 918         helper = __ip_conntrack_helper_find_byname(helpname);
 919         if (!helper) {
 920                 if (!strcmp(helpname, ""))
 921                         helper = NULL;
 922                 else
 923                         return -EINVAL;
 924         }
 925
 926         if (ct->helper) {
 927                 if (!helper) {
 928                         /* we had a helper before ... */
 929                         ip_ct_remove_expectations(ct);
 930                         ct->helper = NULL;
 931                 } else {
 932                         /* need to zero data of old helper */
 933                         memset(&ct->help, 0, sizeof(ct->help));
 934                 }
 935         }
 936
 937         ct->helper = helper;
 938
 939         return 0;
 940 }
 941
 942 static inline int
 943 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 944 {
 945         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 946
 947         if (!del_timer(&ct->timeout))
 948                 return -ETIME;
 949
 950         ct->timeout.expires = jiffies + timeout * HZ;
 951         add_timer(&ct->timeout);
 952
 953         return 0;
 954 }
 955
 956 static inline int
 957 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 958 {
 959         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 960         struct ip_conntrack_protocol *proto;
 961         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 962         int err = 0;
 963
 964         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 965
 966         proto = ip_conntrack_proto_find_get(npt);
 967
 968         if (proto->from_nfattr)
 969                 err = proto->from_nfattr(tb, ct);
 970         ip_conntrack_proto_put(proto);
 971
 972         return err;
 973 }
 974
 975 static int
 976 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 977 {
 978         int err;
 979
 980         DEBUGP("entered %s\n", __FUNCTION__);
 981
 982         if (cda[CTA_HELP-1]) {
 983                 err = ctnetlink_change_helper(ct, cda);
 984                 if (err < 0)
 985                         return err;
 986         }
 987
 988         if (cda[CTA_TIMEOUT-1]) {
 989                 err = ctnetlink_change_timeout(ct, cda);
 990                 if (err < 0)
 991                         return err;
 992         }
 993
 994         if (cda[CTA_STATUS-1]) {
 995                 err = ctnetlink_change_status(ct, cda);
 996                 if (err < 0)
 997                         return err;
 998         }
 999
1000         if (cda[CTA_PROTOINFO-1]) {
1001                 err = ctnetlink_change_protoinfo(ct, cda);
1002                 if (err < 0)
1003                         return err;
1004         }
1005
1006 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1007         if (cda[CTA_MARK-1])
1008                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1009 #endif
1010
1011         DEBUGP("all done\n");
1012         return 0;
1013 }
1014
1015 static int
1016 ctnetlink_create_conntrack(struct nfattr *cda[],
1017                            struct ip_conntrack_tuple *otuple,
1018                            struct ip_conntrack_tuple *rtuple)
1019 {
1020         struct ip_conntrack *ct;
1021         int err = -EINVAL;
1022
1023         DEBUGP("entered %s\n", __FUNCTION__);
1024
1025         ct = ip_conntrack_alloc(otuple, rtuple);
1026         if (ct == NULL || IS_ERR(ct))
1027                 return -ENOMEM;
1028
1029         if (!cda[CTA_TIMEOUT-1])
1030                 goto err;
1031         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1032
1033         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1034         ct->status |= IPS_CONFIRMED;
1035
1036         err = ctnetlink_change_status(ct, cda);
1037         if (err < 0)
1038                 goto err;
1039
1040         if (cda[CTA_PROTOINFO-1]) {
1041                 err = ctnetlink_change_protoinfo(ct, cda);
1042                 if (err < 0)
1043                         return err;
1044         }
1045
1046 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1047         if (cda[CTA_MARK-1])
1048                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1049 #endif
1050
1051         ct->helper = ip_conntrack_helper_find_get(rtuple);
1052
1053         add_timer(&ct->timeout);
1054         ip_conntrack_hash_insert(ct);
1055
1056         if (ct->helper)
1057                 ip_conntrack_helper_put(ct->helper);
1058
1059         DEBUGP("conntrack with id %u inserted\n", ct->id);
1060         return 0;
1061
1062 err:
1063         ip_conntrack_free(ct);
1064         return err;
1065 }
1066
1067 static int
1068 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1069                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1070 {
1071         struct ip_conntrack_tuple otuple, rtuple;
1072         struct ip_conntrack_tuple_hash *h = NULL;
1073         int err = 0;
1074
1075         DEBUGP("entered %s\n", __FUNCTION__);
1076
1077         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1078                 return -EINVAL;
1079
1080         if (cda[CTA_TUPLE_ORIG-1]) {
1081                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1082                 if (err < 0)
1083                         return err;
1084         }
1085
1086         if (cda[CTA_TUPLE_REPLY-1]) {
1087                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1088                 if (err < 0)
1089                         return err;
1090         }
1091
1092         write_lock_bh(&ip_conntrack_lock);
1093         if (cda[CTA_TUPLE_ORIG-1])
1094                 h = __ip_conntrack_find(&otuple, NULL);
1095         else if (cda[CTA_TUPLE_REPLY-1])
1096                 h = __ip_conntrack_find(&rtuple, NULL);
1097
1098         if (h == NULL) {
1099                 write_unlock_bh(&ip_conntrack_lock);
1100                 DEBUGP("no such conntrack, create new\n");
1101                 err = -ENOENT;
1102                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1103                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1104                 return err;
1105         }
1106         /* implicit 'else' */
1107
1108         /* we only allow nat config for new conntracks */
1109         if (cda[CTA_NAT-1]) {
1110                 err = -EINVAL;
1111                 goto out_unlock;
1112         }
1113
1114         /* We manipulate the conntrack inside the global conntrack table lock,
1115          * so there's no need to increase the refcount */
1116         DEBUGP("conntrack found\n");
1117         err = -EEXIST;
1118         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1119                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1120
1121 out_unlock:
1122         write_unlock_bh(&ip_conntrack_lock);
1123         return err;
1124 }
1125
1126 /***********************************************************************
1127  * EXPECT
1128  ***********************************************************************/
1129
1130 static inline int
1131 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1132                          const struct ip_conntrack_tuple *tuple,
1133                          enum ctattr_expect type)
1134 {
1135         struct nfattr *nest_parms = NFA_NEST(skb, type);
1136
1137         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1138                 goto nfattr_failure;
1139
1140         NFA_NEST_END(skb, nest_parms);
1141
1142         return 0;
1143
1144 nfattr_failure:
1145         return -1;
1146 }
1147
1148 static inline int
1149 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1150                         const struct ip_conntrack_tuple *tuple,
1151                         const struct ip_conntrack_tuple *mask)
1152 {
1153         int ret;
1154         struct ip_conntrack_protocol *proto;
1155         struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1156
1157         ret = ctnetlink_dump_tuples_ip(skb, mask);
1158         if (unlikely(ret < 0))
1159                 goto nfattr_failure;
1160
1161         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1162         ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1163         ip_conntrack_proto_put(proto);
1164         if (unlikely(ret < 0))
1165                 goto nfattr_failure;
1166
1167         NFA_NEST_END(skb, nest_parms);
1168
1169         return 0;
1170
1171 nfattr_failure:
1172         return -1;
1173 }
1174
1175 static inline int
1176 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1177                           const struct ip_conntrack_expect *exp)
1178 {
1179         struct ip_conntrack *master = exp->master;
1180         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1181         u_int32_t id = htonl(exp->id);
1182
1183         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1184                 goto nfattr_failure;
1185         if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1186                 goto nfattr_failure;
1187         if (ctnetlink_exp_dump_tuple(skb,
1188                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1189                                  CTA_EXPECT_MASTER) < 0)
1190                 goto nfattr_failure;
1191
1192         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1193         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1194
1195         return 0;
1196
1197 nfattr_failure:
1198         return -1;
1199 }
1200
1201 static int
1202 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1203                     int event,
1204                     int nowait,
1205                     const struct ip_conntrack_expect *exp)
1206 {
1207         struct nlmsghdr *nlh;
1208         struct nfgenmsg *nfmsg;
1209         unsigned char *b;
1210
1211         b = skb->tail;
1212
1213         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1214         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1215         nfmsg  = NLMSG_DATA(nlh);
1216
1217         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1218         nfmsg->nfgen_family = AF_INET;
1219         nfmsg->version      = NFNETLINK_V0;
1220         nfmsg->res_id       = 0;
1221
1222         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1223                 goto nfattr_failure;
1224
1225         nlh->nlmsg_len = skb->tail - b;
1226         return skb->len;
1227
1228 nlmsg_failure:
1229 nfattr_failure:
1230         skb_trim(skb, b - skb->data);
1231         return -1;
1232 }
1233
1234 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1235 static int ctnetlink_expect_event(struct notifier_block *this,
1236                                   unsigned long events, void *ptr)
1237 {
1238         struct nlmsghdr *nlh;
1239         struct nfgenmsg *nfmsg;
1240         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1241         struct sk_buff *skb;
1242         unsigned int type;
1243         unsigned char *b;
1244         int flags = 0;
1245
1246         if (events & IPEXP_NEW) {
1247                 type = IPCTNL_MSG_EXP_NEW;
1248                 flags = NLM_F_CREATE|NLM_F_EXCL;
1249         } else
1250                 return NOTIFY_DONE;
1251
1252         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1253         if (!skb)
1254                 return NOTIFY_DONE;
1255
1256         b = skb->tail;
1257
1258         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1259         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1260         nfmsg = NLMSG_DATA(nlh);
1261
1262         nlh->nlmsg_flags    = flags;
1263         nfmsg->nfgen_family = AF_INET;
1264         nfmsg->version      = NFNETLINK_V0;
1265         nfmsg->res_id       = 0;
1266
1267         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1268                 goto nfattr_failure;
1269
1270         nlh->nlmsg_len = skb->tail - b;
1271         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1272         return NOTIFY_DONE;
1273
1274 nlmsg_failure:
1275 nfattr_failure:
1276         kfree_skb(skb);
1277         return NOTIFY_DONE;
1278 }
1279 #endif
1280
1281 static int
1282 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1283 {
1284         struct ip_conntrack_expect *exp = NULL;
1285         struct list_head *i;
1286         u_int32_t *id = (u_int32_t *) &cb->args[0];
1287
1288         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1289
1290         read_lock_bh(&ip_conntrack_lock);
1291         list_for_each_prev(i, &ip_conntrack_expect_list) {
1292                 exp = (struct ip_conntrack_expect *) i;
1293                 if (exp->id <= *id)
1294                         continue;
1295                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1296                                             cb->nlh->nlmsg_seq,
1297                                             IPCTNL_MSG_EXP_NEW,
1298                                             1, exp) < 0)
1299                         goto out;
1300                 *id = exp->id;
1301         }
1302 out:
1303         read_unlock_bh(&ip_conntrack_lock);
1304
1305         DEBUGP("leaving, last id=%llu\n", *id);
1306
1307         return skb->len;
1308 }
1309
1310 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1311         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1312         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1313 };
1314
1315 static int
1316 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1317                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1318 {
1319         struct ip_conntrack_tuple tuple;
1320         struct ip_conntrack_expect *exp;
1321         struct sk_buff *skb2;
1322         int err = 0;
1323
1324         DEBUGP("entered %s\n", __FUNCTION__);
1325
1326         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1327                 return -EINVAL;
1328
1329         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1330                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1331                 u32 rlen;
1332
1333                 if (msg->nfgen_family != AF_INET)
1334                         return -EAFNOSUPPORT;
1335
1336                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1337                                                 ctnetlink_exp_dump_table,
1338                                                 ctnetlink_done)) != 0)
1339                         return -EINVAL;
1340                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1341                 if (rlen > skb->len)
1342                         rlen = skb->len;
1343                 skb_pull(skb, rlen);
1344                 return 0;
1345         }
1346
1347         if (cda[CTA_EXPECT_MASTER-1])
1348                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1349         else
1350                 return -EINVAL;
1351
1352         if (err < 0)
1353                 return err;
1354
1355         exp = ip_conntrack_expect_find(&tuple);
1356         if (!exp)
1357                 return -ENOENT;
1358
1359         if (cda[CTA_EXPECT_ID-1]) {
1360                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1361                 if (exp->id != ntohl(id)) {
1362                         ip_conntrack_expect_put(exp);
1363                         return -ENOENT;
1364                 }
1365         }
1366
1367         err = -ENOMEM;
1368         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1369         if (!skb2)
1370                 goto out;
1371         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1372
1373         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1374                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1375                                       1, exp);
1376         if (err <= 0)
1377                 goto free;
1378
1379         ip_conntrack_expect_put(exp);
1380
1381         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1382
1383 free:
1384         kfree_skb(skb2);
1385 out:
1386         ip_conntrack_expect_put(exp);
1387         return err;
1388 }
1389
1390 static int
1391 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1392                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1393 {
1394         struct ip_conntrack_expect *exp, *tmp;
1395         struct ip_conntrack_tuple tuple;
1396         struct ip_conntrack_helper *h;
1397         int err;
1398
1399         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1400                 return -EINVAL;
1401
1402         if (cda[CTA_EXPECT_TUPLE-1]) {
1403                 /* delete a single expect by tuple */
1404                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1405                 if (err < 0)
1406                         return err;
1407
1408                 /* bump usage count to 2 */
1409                 exp = ip_conntrack_expect_find(&tuple);
1410                 if (!exp)
1411                         return -ENOENT;
1412
1413                 if (cda[CTA_EXPECT_ID-1]) {
1414                         u_int32_t id =
1415                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1416                         if (exp->id != ntohl(id)) {
1417                                 ip_conntrack_expect_put(exp);
1418                                 return -ENOENT;
1419                         }
1420                 }
1421
1422                 /* after list removal, usage count == 1 */
1423                 ip_conntrack_unexpect_related(exp);
1424                 /* have to put what we 'get' above.
1425                  * after this line usage count == 0 */
1426                 ip_conntrack_expect_put(exp);
1427         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1428                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1429
1430                 /* delete all expectations for this helper */
1431                 write_lock_bh(&ip_conntrack_lock);
1432                 h = __ip_conntrack_helper_find_byname(name);
1433                 if (!h) {
1434                         write_unlock_bh(&ip_conntrack_lock);
1435                         return -EINVAL;
1436                 }
1437                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1438                                          list) {
1439                         if (exp->master->helper == h
1440                             && del_timer(&exp->timeout)) {
1441                                 ip_ct_unlink_expect(exp);
1442                                 ip_conntrack_expect_put(exp);
1443                         }
1444                 }
1445                 write_unlock_bh(&ip_conntrack_lock);
1446         } else {
1447                 /* This basically means we have to flush everything*/
1448                 write_lock_bh(&ip_conntrack_lock);
1449                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1450                                          list) {
1451                         if (del_timer(&exp->timeout)) {
1452                                 ip_ct_unlink_expect(exp);
1453                                 ip_conntrack_expect_put(exp);
1454                         }
1455                 }
1456                 write_unlock_bh(&ip_conntrack_lock);
1457         }
1458
1459         return 0;
1460 }
1461 static int
1462 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1463 {
1464         return -EOPNOTSUPP;
1465 }
1466
1467 static int
1468 ctnetlink_create_expect(struct nfattr *cda[])
1469 {
1470         struct ip_conntrack_tuple tuple, mask, master_tuple;
1471         struct ip_conntrack_tuple_hash *h = NULL;
1472         struct ip_conntrack_expect *exp;
1473         struct ip_conntrack *ct;
1474         int err = 0;
1475
1476         DEBUGP("entered %s\n", __FUNCTION__);
1477
1478         /* caller guarantees that those three CTA_EXPECT_* exist */
1479         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1480         if (err < 0)
1481                 return err;
1482         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1483         if (err < 0)
1484                 return err;
1485         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1486         if (err < 0)
1487                 return err;
1488
1489         /* Look for master conntrack of this expectation */
1490         h = ip_conntrack_find_get(&master_tuple, NULL);
1491         if (!h)
1492                 return -ENOENT;
1493         ct = tuplehash_to_ctrack(h);
1494
1495         if (!ct->helper) {
1496                 /* such conntrack hasn't got any helper, abort */
1497                 err = -EINVAL;
1498                 goto out;
1499         }
1500
1501         exp = ip_conntrack_expect_alloc(ct);
1502         if (!exp) {
1503                 err = -ENOMEM;
1504                 goto out;
1505         }
1506
1507         exp->expectfn = NULL;
1508         exp->flags = 0;
1509         exp->master = ct;
1510         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1511         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1512
1513         err = ip_conntrack_expect_related(exp);
1514         ip_conntrack_expect_put(exp);
1515
1516 out:
1517         ip_conntrack_put(tuplehash_to_ctrack(h));
1518         return err;
1519 }
1520
1521 static int
1522 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1523                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1524 {
1525         struct ip_conntrack_tuple tuple;
1526         struct ip_conntrack_expect *exp;
1527         int err = 0;
1528
1529         DEBUGP("entered %s\n", __FUNCTION__);
1530
1531         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1532                 return -EINVAL;
1533
1534         if (!cda[CTA_EXPECT_TUPLE-1]
1535             || !cda[CTA_EXPECT_MASK-1]
1536             || !cda[CTA_EXPECT_MASTER-1])
1537                 return -EINVAL;
1538
1539         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1540         if (err < 0)
1541                 return err;
1542
1543         write_lock_bh(&ip_conntrack_lock);
1544         exp = __ip_conntrack_expect_find(&tuple);
1545
1546         if (!exp) {
1547                 write_unlock_bh(&ip_conntrack_lock);
1548                 err = -ENOENT;
1549                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1550                         err = ctnetlink_create_expect(cda);
1551                 return err;
1552         }
1553
1554         err = -EEXIST;
1555         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1556                 err = ctnetlink_change_expect(exp, cda);
1557         write_unlock_bh(&ip_conntrack_lock);
1558
1559         DEBUGP("leaving\n");
1560
1561         return err;
1562 }
1563
1564 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1565 static struct notifier_block ctnl_notifier = {
1566         .notifier_call  = ctnetlink_conntrack_event,
1567 };
1568
1569 static struct notifier_block ctnl_notifier_exp = {
1570         .notifier_call  = ctnetlink_expect_event,
1571 };
1572 #endif
1573
1574 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1575         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1576                                             .attr_count = CTA_MAX, },
1577         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1578                                             .attr_count = CTA_MAX, },
1579         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1580                                             .attr_count = CTA_MAX, },
1581         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1582                                             .attr_count = CTA_MAX, },
1583 };
1584
1585 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1586         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1587                                             .attr_count = CTA_EXPECT_MAX, },
1588         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1589                                             .attr_count = CTA_EXPECT_MAX, },
1590         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1591                                             .attr_count = CTA_EXPECT_MAX, },
1592 };
1593
1594 static struct nfnetlink_subsystem ctnl_subsys = {
1595         .name                           = "conntrack",
1596         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1597         .cb_count                       = IPCTNL_MSG_MAX,
1598         .cb                             = ctnl_cb,
1599 };
1600
1601 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1602         .name                           = "conntrack_expect",
1603         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1604         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1605         .cb                             = ctnl_exp_cb,
1606 };
1607
1608 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1609 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1610
1611 static int __init ctnetlink_init(void)
1612 {
1613         int ret;
1614
1615         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1616         ret = nfnetlink_subsys_register(&ctnl_subsys);
1617         if (ret < 0) {
1618                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1619                 goto err_out;
1620         }
1621
1622         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1623         if (ret < 0) {
1624                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1625                 goto err_unreg_subsys;
1626         }
1627
1628 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1629         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1630         if (ret < 0) {
1631                 printk("ctnetlink_init: cannot register notifier.\n");
1632                 goto err_unreg_exp_subsys;
1633         }
1634
1635         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1636         if (ret < 0) {
1637                 printk("ctnetlink_init: cannot expect register notifier.\n");
1638                 goto err_unreg_notifier;
1639         }
1640 #endif
1641
1642         return 0;
1643
1644 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1645 err_unreg_notifier:
1646         ip_conntrack_unregister_notifier(&ctnl_notifier);
1647 err_unreg_exp_subsys:
1648         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1649 #endif
1650 err_unreg_subsys:
1651         nfnetlink_subsys_unregister(&ctnl_subsys);
1652 err_out:
1653         return ret;
1654 }
1655
1656 static void __exit ctnetlink_exit(void)
1657 {
1658         printk("ctnetlink: unregistering from nfnetlink.\n");
1659
1660 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1661         ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1662         ip_conntrack_unregister_notifier(&ctnl_notifier);
1663 #endif
1664
1665         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1666         nfnetlink_subsys_unregister(&ctnl_subsys);
1667         return;
1668 }
1669
1670 module_init(ctnetlink_init);
1671 module_exit(ctnetlink_exit);