net/ipv4/netfilter/ip_conntrack_netlink.c

   1 /* Connection tracking via netlink socket. Allows for user space
   2  * protocol helpers and general trouble making from userspace.
   3  *
   4  * (C) 2001 by Jay Schulist <jschlst@samba.org>
   5  * (C) 2002-2005 by Harald Welte <laforge@gnumonks.org>
   6  * (C) 2003 by Patrick Mchardy <kaber@trash.net>
   7  * (C) 2005 by Pablo Neira Ayuso <pablo@eurodev.net>
   8  *
   9  * I've reworked this stuff to use attributes instead of conntrack
  10  * structures. 5.44 am. I need more tea. --pablo 05/07/11.
  11  *
  12  * Initial connection tracking via netlink development funded and
  13  * generally made possible by Network Robots, Inc. (www.networkrobots.com)
  14  *
  15  * Further development of this code funded by Astaro AG (http://www.astaro.com)
  16  *
  17  * This software may be used and distributed according to the terms
  18  * of the GNU General Public License, incorporated herein by reference.
  19  */
  20
  21 #include <linux/init.h>
  22 #include <linux/module.h>
  23 #include <linux/kernel.h>
  24 #include <linux/types.h>
  25 #include <linux/timer.h>
  26 #include <linux/skbuff.h>
  27 #include <linux/errno.h>
  28 #include <linux/netlink.h>
  29 #include <linux/spinlock.h>
  30 #include <linux/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 {
  58         struct ip_conntrack_protocol *proto;
  59         int ret = 0;
  60
  61         NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
  62
  63         /* If no protocol helper is found, this function will return the
  64          * generic protocol helper, so proto won't *ever* be NULL */
  65         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
  66         if (likely(proto->tuple_to_nfattr))
  67                 ret = proto->tuple_to_nfattr(skb, tuple);
  68
  69         ip_conntrack_proto_put(proto);
  70
  71         return ret;
  72
  73 nfattr_failure:
  74         return -1;
  75 }
  76
  77 static inline int
  78 ctnetlink_dump_tuples(struct sk_buff *skb,
  79                       const struct ip_conntrack_tuple *tuple)
  80 {
  81         struct nfattr *nest_parms;
  82         int ret;
  83
  84         nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
  85         NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
  86         NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
  87         NFA_NEST_END(skb, nest_parms);
  88
  89         nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
  90         ret = ctnetlink_dump_tuples_proto(skb, tuple);
  91         NFA_NEST_END(skb, nest_parms);
  92
  93         return ret;
  94
  95 nfattr_failure:
  96         return -1;
  97 }
  98
  99 static inline int
 100 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
 101 {
 102         u_int32_t status = htonl((u_int32_t) ct->status);
 103         NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
 104         return 0;
 105
 106 nfattr_failure:
 107         return -1;
 108 }
 109
 110 static inline int
 111 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
 112 {
 113         long timeout_l = ct->timeout.expires - jiffies;
 114         u_int32_t timeout;
 115
 116         if (timeout_l < 0)
 117                 timeout = 0;
 118         else
 119                 timeout = htonl(timeout_l / HZ);
 120
 121         NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
 122         return 0;
 123
 124 nfattr_failure:
 125         return -1;
 126 }
 127
 128 static inline int
 129 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 130 {
 131         struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 132
 133         struct nfattr *nest_proto;
 134         int ret;
 135
 136         if (!proto->to_nfattr) {
 137                 ip_conntrack_proto_put(proto);
 138                 return 0;
 139         }
 140
 141         nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
 142
 143         ret = proto->to_nfattr(skb, nest_proto, ct);
 144
 145         ip_conntrack_proto_put(proto);
 146
 147         NFA_NEST_END(skb, nest_proto);
 148
 149         return ret;
 150
 151 nfattr_failure:
 152         return -1;
 153 }
 154
 155 static inline int
 156 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
 157 {
 158         struct nfattr *nest_helper;
 159
 160         if (!ct->helper)
 161                 return 0;
 162
 163         nest_helper = NFA_NEST(skb, CTA_HELP);
 164         NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
 165
 166         if (ct->helper->to_nfattr)
 167                 ct->helper->to_nfattr(skb, ct);
 168
 169         NFA_NEST_END(skb, nest_helper);
 170
 171         return 0;
 172
 173 nfattr_failure:
 174         return -1;
 175 }
 176
 177 #ifdef CONFIG_IP_NF_CT_ACCT
 178 static inline int
 179 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
 180                         enum ip_conntrack_dir dir)
 181 {
 182         enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
 183         struct nfattr *nest_count = NFA_NEST(skb, type);
 184         u_int32_t tmp;
 185
 186         tmp = htonl(ct->counters[dir].packets);
 187         NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
 188
 189         tmp = htonl(ct->counters[dir].bytes);
 190         NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
 191
 192         NFA_NEST_END(skb, nest_count);
 193
 194         return 0;
 195
 196 nfattr_failure:
 197         return -1;
 198 }
 199 #else
 200 #define ctnetlink_dump_counters(a, b, c) (0)
 201 #endif
 202
 203 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
 204 static inline int
 205 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
 206 {
 207         u_int32_t mark = htonl(ct->mark);
 208
 209         NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
 210         return 0;
 211
 212 nfattr_failure:
 213         return -1;
 214 }
 215 #else
 216 #define ctnetlink_dump_mark(a, b) (0)
 217 #endif
 218
 219 static inline int
 220 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
 221 {
 222         u_int32_t id = htonl(ct->id);
 223         NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
 224         return 0;
 225
 226 nfattr_failure:
 227         return -1;
 228 }
 229
 230 static inline int
 231 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
 232 {
 233         u_int32_t use = htonl(atomic_read(&ct->ct_general.use));
 234
 235         NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
 236         return 0;
 237
 238 nfattr_failure:
 239         return -1;
 240 }
 241
 242 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
 243
 244 static int
 245 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
 246                     int event, int nowait,
 247                     const struct ip_conntrack *ct)
 248 {
 249         struct nlmsghdr *nlh;
 250         struct nfgenmsg *nfmsg;
 251         struct nfattr *nest_parms;
 252         unsigned char *b;
 253
 254         b = skb->tail;
 255
 256         event |= NFNL_SUBSYS_CTNETLINK << 8;
 257         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
 258         nfmsg  = NLMSG_DATA(nlh);
 259
 260         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
 261         nfmsg->nfgen_family = AF_INET;
 262         nfmsg->version      = NFNETLINK_V0;
 263         nfmsg->res_id       = 0;
 264
 265         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 266         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 267                 goto nfattr_failure;
 268         NFA_NEST_END(skb, nest_parms);
 269
 270         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 271         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 272                 goto nfattr_failure;
 273         NFA_NEST_END(skb, nest_parms);
 274
 275         if (ctnetlink_dump_status(skb, ct) < 0 ||
 276             ctnetlink_dump_timeout(skb, ct) < 0 ||
 277             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 278             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
 279             ctnetlink_dump_protoinfo(skb, ct) < 0 ||
 280             ctnetlink_dump_helpinfo(skb, ct) < 0 ||
 281             ctnetlink_dump_mark(skb, ct) < 0 ||
 282             ctnetlink_dump_id(skb, ct) < 0 ||
 283             ctnetlink_dump_use(skb, ct) < 0)
 284                 goto nfattr_failure;
 285
 286         nlh->nlmsg_len = skb->tail - b;
 287         return skb->len;
 288
 289 nlmsg_failure:
 290 nfattr_failure:
 291         skb_trim(skb, b - skb->data);
 292         return -1;
 293 }
 294
 295 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
 296 static int ctnetlink_conntrack_event(struct notifier_block *this,
 297                                      unsigned long events, void *ptr)
 298 {
 299         struct nlmsghdr *nlh;
 300         struct nfgenmsg *nfmsg;
 301         struct nfattr *nest_parms;
 302         struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
 303         struct sk_buff *skb;
 304         unsigned int type;
 305         unsigned char *b;
 306         unsigned int flags = 0, group;
 307
 308         /* ignore our fake conntrack entry */
 309         if (ct == &ip_conntrack_untracked)
 310                 return NOTIFY_DONE;
 311
 312         if (events & IPCT_DESTROY) {
 313                 type = IPCTNL_MSG_CT_DELETE;
 314                 group = NFNLGRP_CONNTRACK_DESTROY;
 315         } else if (events & (IPCT_NEW | IPCT_RELATED)) {
 316                 type = IPCTNL_MSG_CT_NEW;
 317                 flags = NLM_F_CREATE|NLM_F_EXCL;
 318                 /* dump everything */
 319                 events = ~0UL;
 320                 group = NFNLGRP_CONNTRACK_NEW;
 321         } else if (events & (IPCT_STATUS |
 322                       IPCT_PROTOINFO |
 323                       IPCT_HELPER |
 324                       IPCT_HELPINFO |
 325                       IPCT_NATINFO)) {
 326                 type = IPCTNL_MSG_CT_NEW;
 327                 group = NFNLGRP_CONNTRACK_UPDATE;
 328         } else
 329                 return NOTIFY_DONE;
 330
 331   /* FIXME: Check if there are any listeners before, don't hurt performance */
 332
 333         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
 334         if (!skb)
 335                 return NOTIFY_DONE;
 336
 337         b = skb->tail;
 338
 339         type |= NFNL_SUBSYS_CTNETLINK << 8;
 340         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
 341         nfmsg = NLMSG_DATA(nlh);
 342
 343         nlh->nlmsg_flags    = flags;
 344         nfmsg->nfgen_family = AF_INET;
 345         nfmsg->version  = NFNETLINK_V0;
 346         nfmsg->res_id   = 0;
 347
 348         nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
 349         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
 350                 goto nfattr_failure;
 351         NFA_NEST_END(skb, nest_parms);
 352
 353         nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
 354         if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
 355                 goto nfattr_failure;
 356         NFA_NEST_END(skb, nest_parms);
 357
 358         /* NAT stuff is now a status flag */
 359         if ((events & IPCT_STATUS || events & IPCT_NATINFO)
 360             && ctnetlink_dump_status(skb, ct) < 0)
 361                 goto nfattr_failure;
 362         if (events & IPCT_REFRESH
 363             && ctnetlink_dump_timeout(skb, ct) < 0)
 364                 goto nfattr_failure;
 365         if (events & IPCT_PROTOINFO
 366             && ctnetlink_dump_protoinfo(skb, ct) < 0)
 367                 goto nfattr_failure;
 368         if (events & IPCT_HELPINFO
 369             && ctnetlink_dump_helpinfo(skb, ct) < 0)
 370                 goto nfattr_failure;
 371
 372         if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
 373             ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
 374                 goto nfattr_failure;
 375
 376         nlh->nlmsg_len = skb->tail - b;
 377         nfnetlink_send(skb, 0, group, 0);
 378         return NOTIFY_DONE;
 379
 380 nlmsg_failure:
 381 nfattr_failure:
 382         kfree_skb(skb);
 383         return NOTIFY_DONE;
 384 }
 385 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
 386
 387 static int ctnetlink_done(struct netlink_callback *cb)
 388 {
 389         DEBUGP("entered %s\n", __FUNCTION__);
 390         return 0;
 391 }
 392
 393 static int
 394 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
 395 {
 396         struct ip_conntrack *ct = NULL;
 397         struct ip_conntrack_tuple_hash *h;
 398         struct list_head *i;
 399         u_int32_t *id = (u_int32_t *) &cb->args[1];
 400
 401         DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
 402                         cb->args[0], *id);
 403
 404         read_lock_bh(&ip_conntrack_lock);
 405         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 406                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 407                         h = (struct ip_conntrack_tuple_hash *) i;
 408                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 409                                 continue;
 410                         ct = tuplehash_to_ctrack(h);
 411                         if (ct->id <= *id)
 412                                 continue;
 413                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 414                                                 cb->nlh->nlmsg_seq,
 415                                                 IPCTNL_MSG_CT_NEW,
 416                                                 1, ct) < 0)
 417                                 goto out;
 418                         *id = ct->id;
 419                 }
 420         }
 421 out:
 422         read_unlock_bh(&ip_conntrack_lock);
 423
 424         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 425
 426         return skb->len;
 427 }
 428
 429 #ifdef CONFIG_IP_NF_CT_ACCT
 430 static int
 431 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
 432 {
 433         struct ip_conntrack *ct = NULL;
 434         struct ip_conntrack_tuple_hash *h;
 435         struct list_head *i;
 436         u_int32_t *id = (u_int32_t *) &cb->args[1];
 437
 438         DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
 439                         cb->args[0], *id);
 440
 441         write_lock_bh(&ip_conntrack_lock);
 442         for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
 443                 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
 444                         h = (struct ip_conntrack_tuple_hash *) i;
 445                         if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
 446                                 continue;
 447                         ct = tuplehash_to_ctrack(h);
 448                         if (ct->id <= *id)
 449                                 continue;
 450                         if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
 451                                                 cb->nlh->nlmsg_seq,
 452                                                 IPCTNL_MSG_CT_NEW,
 453                                                 1, ct) < 0)
 454                                 goto out;
 455                         *id = ct->id;
 456
 457                         memset(&ct->counters, 0, sizeof(ct->counters));
 458                 }
 459         }
 460 out:
 461         write_unlock_bh(&ip_conntrack_lock);
 462
 463         DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
 464
 465         return skb->len;
 466 }
 467 #endif
 468
 469 static const size_t cta_min_ip[CTA_IP_MAX] = {
 470         [CTA_IP_V4_SRC-1]       = sizeof(u_int32_t),
 471         [CTA_IP_V4_DST-1]       = sizeof(u_int32_t),
 472 };
 473
 474 static inline int
 475 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
 476 {
 477         struct nfattr *tb[CTA_IP_MAX];
 478
 479         DEBUGP("entered %s\n", __FUNCTION__);
 480
 481         nfattr_parse_nested(tb, CTA_IP_MAX, attr);
 482
 483         if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
 484                 return -EINVAL;
 485
 486         if (!tb[CTA_IP_V4_SRC-1])
 487                 return -EINVAL;
 488         tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
 489
 490         if (!tb[CTA_IP_V4_DST-1])
 491                 return -EINVAL;
 492         tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
 493
 494         DEBUGP("leaving\n");
 495
 496         return 0;
 497 }
 498
 499 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
 500         [CTA_PROTO_NUM-1]       = sizeof(u_int8_t),
 501         [CTA_PROTO_SRC_PORT-1]  = sizeof(u_int16_t),
 502         [CTA_PROTO_DST_PORT-1]  = sizeof(u_int16_t),
 503         [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
 504         [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
 505         [CTA_PROTO_ICMP_ID-1]   = sizeof(u_int16_t),
 506 };
 507
 508 static inline int
 509 ctnetlink_parse_tuple_proto(struct nfattr *attr,
 510                             struct ip_conntrack_tuple *tuple)
 511 {
 512         struct nfattr *tb[CTA_PROTO_MAX];
 513         struct ip_conntrack_protocol *proto;
 514         int ret = 0;
 515
 516         DEBUGP("entered %s\n", __FUNCTION__);
 517
 518         nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
 519
 520         if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
 521                 return -EINVAL;
 522
 523         if (!tb[CTA_PROTO_NUM-1])
 524                 return -EINVAL;
 525         tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
 526
 527         proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
 528
 529         if (likely(proto->nfattr_to_tuple))
 530                 ret = proto->nfattr_to_tuple(tb, tuple);
 531
 532         ip_conntrack_proto_put(proto);
 533
 534         return ret;
 535 }
 536
 537 static inline int
 538 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
 539                       enum ctattr_tuple type)
 540 {
 541         struct nfattr *tb[CTA_TUPLE_MAX];
 542         int err;
 543
 544         DEBUGP("entered %s\n", __FUNCTION__);
 545
 546         memset(tuple, 0, sizeof(*tuple));
 547
 548         nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
 549
 550         if (!tb[CTA_TUPLE_IP-1])
 551                 return -EINVAL;
 552
 553         err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
 554         if (err < 0)
 555                 return err;
 556
 557         if (!tb[CTA_TUPLE_PROTO-1])
 558                 return -EINVAL;
 559
 560         err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
 561         if (err < 0)
 562                 return err;
 563
 564         /* orig and expect tuples get DIR_ORIGINAL */
 565         if (type == CTA_TUPLE_REPLY)
 566                 tuple->dst.dir = IP_CT_DIR_REPLY;
 567         else
 568                 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
 569
 570         DUMP_TUPLE(tuple);
 571
 572         DEBUGP("leaving\n");
 573
 574         return 0;
 575 }
 576
 577 #ifdef CONFIG_IP_NF_NAT_NEEDED
 578 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
 579         [CTA_PROTONAT_PORT_MIN-1]       = sizeof(u_int16_t),
 580         [CTA_PROTONAT_PORT_MAX-1]       = sizeof(u_int16_t),
 581 };
 582
 583 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
 584                                      const struct ip_conntrack *ct,
 585                                      struct ip_nat_range *range)
 586 {
 587         struct nfattr *tb[CTA_PROTONAT_MAX];
 588         struct ip_nat_protocol *npt;
 589
 590         DEBUGP("entered %s\n", __FUNCTION__);
 591
 592         nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
 593
 594         if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
 595                 return -EINVAL;
 596
 597         npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
 598
 599         if (!npt->nfattr_to_range) {
 600                 ip_nat_proto_put(npt);
 601                 return 0;
 602         }
 603
 604         /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
 605         if (npt->nfattr_to_range(tb, range) > 0)
 606                 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
 607
 608         ip_nat_proto_put(npt);
 609
 610         DEBUGP("leaving\n");
 611         return 0;
 612 }
 613
 614 static const size_t cta_min_nat[CTA_NAT_MAX] = {
 615         [CTA_NAT_MINIP-1]       = sizeof(u_int32_t),
 616         [CTA_NAT_MAXIP-1]       = sizeof(u_int32_t),
 617 };
 618
 619 static inline int
 620 ctnetlink_parse_nat(struct nfattr *cda[],
 621                     const struct ip_conntrack *ct, struct ip_nat_range *range)
 622 {
 623         struct nfattr *tb[CTA_NAT_MAX];
 624         int err;
 625
 626         DEBUGP("entered %s\n", __FUNCTION__);
 627
 628         memset(range, 0, sizeof(*range));
 629
 630         nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
 631
 632         if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
 633                 return -EINVAL;
 634
 635         if (tb[CTA_NAT_MINIP-1])
 636                 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
 637
 638         if (!tb[CTA_NAT_MAXIP-1])
 639                 range->max_ip = range->min_ip;
 640         else
 641                 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
 642
 643         if (range->min_ip)
 644                 range->flags |= IP_NAT_RANGE_MAP_IPS;
 645
 646         if (!tb[CTA_NAT_PROTO-1])
 647                 return 0;
 648
 649         err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
 650         if (err < 0)
 651                 return err;
 652
 653         DEBUGP("leaving\n");
 654         return 0;
 655 }
 656 #endif
 657
 658 static inline int
 659 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
 660 {
 661         struct nfattr *tb[CTA_HELP_MAX];
 662
 663         DEBUGP("entered %s\n", __FUNCTION__);
 664
 665         nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
 666
 667         if (!tb[CTA_HELP_NAME-1])
 668                 return -EINVAL;
 669
 670         *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
 671
 672         return 0;
 673 }
 674
 675 static const size_t cta_min[CTA_MAX] = {
 676         [CTA_STATUS-1]          = sizeof(u_int32_t),
 677         [CTA_TIMEOUT-1]         = sizeof(u_int32_t),
 678         [CTA_MARK-1]            = sizeof(u_int32_t),
 679         [CTA_USE-1]             = sizeof(u_int32_t),
 680         [CTA_ID-1]              = sizeof(u_int32_t)
 681 };
 682
 683 static int
 684 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
 685                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 686 {
 687         struct ip_conntrack_tuple_hash *h;
 688         struct ip_conntrack_tuple tuple;
 689         struct ip_conntrack *ct;
 690         int err = 0;
 691
 692         DEBUGP("entered %s\n", __FUNCTION__);
 693
 694         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 695                 return -EINVAL;
 696
 697         if (cda[CTA_TUPLE_ORIG-1])
 698                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 699         else if (cda[CTA_TUPLE_REPLY-1])
 700                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 701         else {
 702                 /* Flush the whole table */
 703                 ip_conntrack_flush();
 704                 return 0;
 705         }
 706
 707         if (err < 0)
 708                 return err;
 709
 710         h = ip_conntrack_find_get(&tuple, NULL);
 711         if (!h) {
 712                 DEBUGP("tuple not found in conntrack hash\n");
 713                 return -ENOENT;
 714         }
 715
 716         ct = tuplehash_to_ctrack(h);
 717
 718         if (cda[CTA_ID-1]) {
 719                 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
 720                 if (ct->id != id) {
 721                         ip_conntrack_put(ct);
 722                         return -ENOENT;
 723                 }
 724         }
 725         if (del_timer(&ct->timeout))
 726                 ct->timeout.function((unsigned long)ct);
 727
 728         ip_conntrack_put(ct);
 729         DEBUGP("leaving\n");
 730
 731         return 0;
 732 }
 733
 734 static int
 735 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
 736                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
 737 {
 738         struct ip_conntrack_tuple_hash *h;
 739         struct ip_conntrack_tuple tuple;
 740         struct ip_conntrack *ct;
 741         struct sk_buff *skb2 = NULL;
 742         int err = 0;
 743
 744         DEBUGP("entered %s\n", __FUNCTION__);
 745
 746         if (nlh->nlmsg_flags & NLM_F_DUMP) {
 747                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
 748                 u32 rlen;
 749
 750                 if (msg->nfgen_family != AF_INET)
 751                         return -EAFNOSUPPORT;
 752
 753                 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
 754                                         IPCTNL_MSG_CT_GET_CTRZERO) {
 755 #ifdef CONFIG_IP_NF_CT_ACCT
 756                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 757                                                 ctnetlink_dump_table_w,
 758                                                 ctnetlink_done)) != 0)
 759                                 return -EINVAL;
 760 #else
 761                         return -ENOTSUPP;
 762 #endif
 763                 } else {
 764                         if ((*errp = netlink_dump_start(ctnl, skb, nlh,
 765                                                         ctnetlink_dump_table,
 766                                                         ctnetlink_done)) != 0)
 767                         return -EINVAL;
 768                 }
 769
 770                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
 771                 if (rlen > skb->len)
 772                         rlen = skb->len;
 773                 skb_pull(skb, rlen);
 774                 return 0;
 775         }
 776
 777         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
 778                 return -EINVAL;
 779
 780         if (cda[CTA_TUPLE_ORIG-1])
 781                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
 782         else if (cda[CTA_TUPLE_REPLY-1])
 783                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
 784         else
 785                 return -EINVAL;
 786
 787         if (err < 0)
 788                 return err;
 789
 790         h = ip_conntrack_find_get(&tuple, NULL);
 791         if (!h) {
 792                 DEBUGP("tuple not found in conntrack hash");
 793                 return -ENOENT;
 794         }
 795         DEBUGP("tuple found\n");
 796         ct = tuplehash_to_ctrack(h);
 797
 798         err = -ENOMEM;
 799         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
 800         if (!skb2) {
 801                 ip_conntrack_put(ct);
 802                 return -ENOMEM;
 803         }
 804         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
 805
 806         err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
 807                                   IPCTNL_MSG_CT_NEW, 1, ct);
 808         ip_conntrack_put(ct);
 809         if (err <= 0)
 810                 goto free;
 811
 812         err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
 813         if (err < 0)
 814                 goto out;
 815
 816         DEBUGP("leaving\n");
 817         return 0;
 818
 819 free:
 820         kfree_skb(skb2);
 821 out:
 822         return err;
 823 }
 824
 825 static inline int
 826 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
 827 {
 828         unsigned long d;
 829         unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
 830         d = ct->status ^ status;
 831
 832         if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
 833                 /* unchangeable */
 834                 return -EINVAL;
 835
 836         if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
 837                 /* SEEN_REPLY bit can only be set */
 838                 return -EINVAL;
 839
 840
 841         if (d & IPS_ASSURED && !(status & IPS_ASSURED))
 842                 /* ASSURED bit can only be set */
 843                 return -EINVAL;
 844
 845         if (cda[CTA_NAT-1]) {
 846 #ifndef CONFIG_IP_NF_NAT_NEEDED
 847                 return -EINVAL;
 848 #else
 849                 unsigned int hooknum;
 850                 struct ip_nat_range range;
 851
 852                 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
 853                         return -EINVAL;
 854
 855                 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
 856                        NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
 857                        htons(range.min.all), htons(range.max.all));
 858
 859                 /* This is tricky but it works. ip_nat_setup_info needs the
 860                  * hook number as parameter, so let's do the correct
 861                  * conversion and run away */
 862                 if (status & IPS_SRC_NAT_DONE)
 863                         hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
 864                 else if (status & IPS_DST_NAT_DONE)
 865                         hooknum = NF_IP_PRE_ROUTING;  /* IP_NAT_MANIP_DST */
 866                 else
 867                         return -EINVAL; /* Missing NAT flags */
 868
 869                 DEBUGP("NAT status: %lu\n",
 870                        status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 871
 872                 if (ip_nat_initialized(ct, HOOK2MANIP(hooknum)))
 873                         return -EEXIST;
 874                 ip_nat_setup_info(ct, &range, hooknum);
 875
 876                 DEBUGP("NAT status after setup_info: %lu\n",
 877                        ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
 878 #endif
 879         }
 880
 881         /* Be careful here, modifying NAT bits can screw up things,
 882          * so don't let users modify them directly if they don't pass
 883          * ip_nat_range. */
 884         ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
 885         return 0;
 886 }
 887
 888
 889 static inline int
 890 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
 891 {
 892         struct ip_conntrack_helper *helper;
 893         char *helpname;
 894         int err;
 895
 896         DEBUGP("entered %s\n", __FUNCTION__);
 897
 898         /* don't change helper of sibling connections */
 899         if (ct->master)
 900                 return -EINVAL;
 901
 902         err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
 903         if (err < 0)
 904                 return err;
 905
 906         helper = __ip_conntrack_helper_find_byname(helpname);
 907         if (!helper) {
 908                 if (!strcmp(helpname, ""))
 909                         helper = NULL;
 910                 else
 911                         return -EINVAL;
 912         }
 913
 914         if (ct->helper) {
 915                 if (!helper) {
 916                         /* we had a helper before ... */
 917                         ip_ct_remove_expectations(ct);
 918                         ct->helper = NULL;
 919                 } else {
 920                         /* need to zero data of old helper */
 921                         memset(&ct->help, 0, sizeof(ct->help));
 922                 }
 923         }
 924
 925         ct->helper = helper;
 926
 927         return 0;
 928 }
 929
 930 static inline int
 931 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
 932 {
 933         u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
 934
 935         if (!del_timer(&ct->timeout))
 936                 return -ETIME;
 937
 938         ct->timeout.expires = jiffies + timeout * HZ;
 939         add_timer(&ct->timeout);
 940
 941         return 0;
 942 }
 943
 944 static inline int
 945 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
 946 {
 947         struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
 948         struct ip_conntrack_protocol *proto;
 949         u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
 950         int err = 0;
 951
 952         nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
 953
 954         proto = ip_conntrack_proto_find_get(npt);
 955
 956         if (proto->from_nfattr)
 957                 err = proto->from_nfattr(tb, ct);
 958         ip_conntrack_proto_put(proto);
 959
 960         return err;
 961 }
 962
 963 static int
 964 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
 965 {
 966         int err;
 967
 968         DEBUGP("entered %s\n", __FUNCTION__);
 969
 970         if (cda[CTA_HELP-1]) {
 971                 err = ctnetlink_change_helper(ct, cda);
 972                 if (err < 0)
 973                         return err;
 974         }
 975
 976         if (cda[CTA_TIMEOUT-1]) {
 977                 err = ctnetlink_change_timeout(ct, cda);
 978                 if (err < 0)
 979                         return err;
 980         }
 981
 982         if (cda[CTA_STATUS-1]) {
 983                 err = ctnetlink_change_status(ct, cda);
 984                 if (err < 0)
 985                         return err;
 986         }
 987
 988         if (cda[CTA_PROTOINFO-1]) {
 989                 err = ctnetlink_change_protoinfo(ct, cda);
 990                 if (err < 0)
 991                         return err;
 992         }
 993
 994 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
 995         if (cda[CTA_MARK-1])
 996                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
 997 #endif
 998
 999         DEBUGP("all done\n");
1000         return 0;
1001 }
1002
1003 static int
1004 ctnetlink_create_conntrack(struct nfattr *cda[],
1005                            struct ip_conntrack_tuple *otuple,
1006                            struct ip_conntrack_tuple *rtuple)
1007 {
1008         struct ip_conntrack *ct;
1009         int err = -EINVAL;
1010
1011         DEBUGP("entered %s\n", __FUNCTION__);
1012
1013         ct = ip_conntrack_alloc(otuple, rtuple);
1014         if (ct == NULL || IS_ERR(ct))
1015                 return -ENOMEM;
1016
1017         if (!cda[CTA_TIMEOUT-1])
1018                 goto err;
1019         ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1020
1021         ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1022         ct->status |= IPS_CONFIRMED;
1023
1024         err = ctnetlink_change_status(ct, cda);
1025         if (err < 0)
1026                 goto err;
1027
1028         if (cda[CTA_PROTOINFO-1]) {
1029                 err = ctnetlink_change_protoinfo(ct, cda);
1030                 if (err < 0)
1031                         return err;
1032         }
1033
1034 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1035         if (cda[CTA_MARK-1])
1036                 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1037 #endif
1038
1039         ct->helper = ip_conntrack_helper_find_get(rtuple);
1040
1041         add_timer(&ct->timeout);
1042         ip_conntrack_hash_insert(ct);
1043
1044         if (ct->helper)
1045                 ip_conntrack_helper_put(ct->helper);
1046
1047         DEBUGP("conntrack with id %u inserted\n", ct->id);
1048         return 0;
1049
1050 err:
1051         ip_conntrack_free(ct);
1052         return err;
1053 }
1054
1055 static int
1056 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1057                         struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1058 {
1059         struct ip_conntrack_tuple otuple, rtuple;
1060         struct ip_conntrack_tuple_hash *h = NULL;
1061         int err = 0;
1062
1063         DEBUGP("entered %s\n", __FUNCTION__);
1064
1065         if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1066                 return -EINVAL;
1067
1068         if (cda[CTA_TUPLE_ORIG-1]) {
1069                 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1070                 if (err < 0)
1071                         return err;
1072         }
1073
1074         if (cda[CTA_TUPLE_REPLY-1]) {
1075                 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1076                 if (err < 0)
1077                         return err;
1078         }
1079
1080         write_lock_bh(&ip_conntrack_lock);
1081         if (cda[CTA_TUPLE_ORIG-1])
1082                 h = __ip_conntrack_find(&otuple, NULL);
1083         else if (cda[CTA_TUPLE_REPLY-1])
1084                 h = __ip_conntrack_find(&rtuple, NULL);
1085
1086         if (h == NULL) {
1087                 write_unlock_bh(&ip_conntrack_lock);
1088                 DEBUGP("no such conntrack, create new\n");
1089                 err = -ENOENT;
1090                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1091                         err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1092                 return err;
1093         }
1094         /* implicit 'else' */
1095
1096         /* we only allow nat config for new conntracks */
1097         if (cda[CTA_NAT-1]) {
1098                 err = -EINVAL;
1099                 goto out_unlock;
1100         }
1101
1102         /* We manipulate the conntrack inside the global conntrack table lock,
1103          * so there's no need to increase the refcount */
1104         DEBUGP("conntrack found\n");
1105         err = -EEXIST;
1106         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1107                 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1108
1109 out_unlock:
1110         write_unlock_bh(&ip_conntrack_lock);
1111         return err;
1112 }
1113
1114 /***********************************************************************
1115  * EXPECT
1116  ***********************************************************************/
1117
1118 static inline int
1119 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1120                          const struct ip_conntrack_tuple *tuple,
1121                          enum ctattr_expect type)
1122 {
1123         struct nfattr *nest_parms = NFA_NEST(skb, type);
1124
1125         if (ctnetlink_dump_tuples(skb, tuple) < 0)
1126                 goto nfattr_failure;
1127
1128         NFA_NEST_END(skb, nest_parms);
1129
1130         return 0;
1131
1132 nfattr_failure:
1133         return -1;
1134 }
1135
1136 static inline int
1137 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1138                           const struct ip_conntrack_expect *exp)
1139 {
1140         struct ip_conntrack *master = exp->master;
1141         u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1142         u_int32_t id = htonl(exp->id);
1143
1144         if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1145                 goto nfattr_failure;
1146         if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1147                 goto nfattr_failure;
1148         if (ctnetlink_exp_dump_tuple(skb,
1149                                  &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1150                                  CTA_EXPECT_MASTER) < 0)
1151                 goto nfattr_failure;
1152
1153         NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1154         NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1155
1156         return 0;
1157
1158 nfattr_failure:
1159         return -1;
1160 }
1161
1162 static int
1163 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1164                     int event,
1165                     int nowait,
1166                     const struct ip_conntrack_expect *exp)
1167 {
1168         struct nlmsghdr *nlh;
1169         struct nfgenmsg *nfmsg;
1170         unsigned char *b;
1171
1172         b = skb->tail;
1173
1174         event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1175         nlh    = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1176         nfmsg  = NLMSG_DATA(nlh);
1177
1178         nlh->nlmsg_flags    = (nowait && pid) ? NLM_F_MULTI : 0;
1179         nfmsg->nfgen_family = AF_INET;
1180         nfmsg->version      = NFNETLINK_V0;
1181         nfmsg->res_id       = 0;
1182
1183         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1184                 goto nfattr_failure;
1185
1186         nlh->nlmsg_len = skb->tail - b;
1187         return skb->len;
1188
1189 nlmsg_failure:
1190 nfattr_failure:
1191         skb_trim(skb, b - skb->data);
1192         return -1;
1193 }
1194
1195 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1196 static int ctnetlink_expect_event(struct notifier_block *this,
1197                                   unsigned long events, void *ptr)
1198 {
1199         struct nlmsghdr *nlh;
1200         struct nfgenmsg *nfmsg;
1201         struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1202         struct sk_buff *skb;
1203         unsigned int type;
1204         unsigned char *b;
1205         int flags = 0;
1206
1207         if (events & IPEXP_NEW) {
1208                 type = IPCTNL_MSG_EXP_NEW;
1209                 flags = NLM_F_CREATE|NLM_F_EXCL;
1210         } else
1211                 return NOTIFY_DONE;
1212
1213         skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1214         if (!skb)
1215                 return NOTIFY_DONE;
1216
1217         b = skb->tail;
1218
1219         type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1220         nlh   = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1221         nfmsg = NLMSG_DATA(nlh);
1222
1223         nlh->nlmsg_flags    = flags;
1224         nfmsg->nfgen_family = AF_INET;
1225         nfmsg->version      = NFNETLINK_V0;
1226         nfmsg->res_id       = 0;
1227
1228         if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1229                 goto nfattr_failure;
1230
1231         nlh->nlmsg_len = skb->tail - b;
1232         nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1233         return NOTIFY_DONE;
1234
1235 nlmsg_failure:
1236 nfattr_failure:
1237         kfree_skb(skb);
1238         return NOTIFY_DONE;
1239 }
1240 #endif
1241
1242 static int
1243 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1244 {
1245         struct ip_conntrack_expect *exp = NULL;
1246         struct list_head *i;
1247         u_int32_t *id = (u_int32_t *) &cb->args[0];
1248
1249         DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1250
1251         read_lock_bh(&ip_conntrack_lock);
1252         list_for_each_prev(i, &ip_conntrack_expect_list) {
1253                 exp = (struct ip_conntrack_expect *) i;
1254                 if (exp->id <= *id)
1255                         continue;
1256                 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1257                                             cb->nlh->nlmsg_seq,
1258                                             IPCTNL_MSG_EXP_NEW,
1259                                             1, exp) < 0)
1260                         goto out;
1261                 *id = exp->id;
1262         }
1263 out:
1264         read_unlock_bh(&ip_conntrack_lock);
1265
1266         DEBUGP("leaving, last id=%llu\n", *id);
1267
1268         return skb->len;
1269 }
1270
1271 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1272         [CTA_EXPECT_TIMEOUT-1]          = sizeof(u_int32_t),
1273         [CTA_EXPECT_ID-1]               = sizeof(u_int32_t)
1274 };
1275
1276 static int
1277 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1278                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1279 {
1280         struct ip_conntrack_tuple tuple;
1281         struct ip_conntrack_expect *exp;
1282         struct sk_buff *skb2;
1283         int err = 0;
1284
1285         DEBUGP("entered %s\n", __FUNCTION__);
1286
1287         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1288                 return -EINVAL;
1289
1290         if (nlh->nlmsg_flags & NLM_F_DUMP) {
1291                 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1292                 u32 rlen;
1293
1294                 if (msg->nfgen_family != AF_INET)
1295                         return -EAFNOSUPPORT;
1296
1297                 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1298                                                 ctnetlink_exp_dump_table,
1299                                                 ctnetlink_done)) != 0)
1300                         return -EINVAL;
1301                 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1302                 if (rlen > skb->len)
1303                         rlen = skb->len;
1304                 skb_pull(skb, rlen);
1305                 return 0;
1306         }
1307
1308         if (cda[CTA_EXPECT_MASTER-1])
1309                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1310         else
1311                 return -EINVAL;
1312
1313         if (err < 0)
1314                 return err;
1315
1316         exp = ip_conntrack_expect_find(&tuple);
1317         if (!exp)
1318                 return -ENOENT;
1319
1320         if (cda[CTA_EXPECT_ID-1]) {
1321                 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1322                 if (exp->id != ntohl(id)) {
1323                         ip_conntrack_expect_put(exp);
1324                         return -ENOENT;
1325                 }
1326         }
1327
1328         err = -ENOMEM;
1329         skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1330         if (!skb2)
1331                 goto out;
1332         NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1333
1334         err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1335                                       nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1336                                       1, exp);
1337         if (err <= 0)
1338                 goto free;
1339
1340         ip_conntrack_expect_put(exp);
1341
1342         return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1343
1344 free:
1345         kfree_skb(skb2);
1346 out:
1347         ip_conntrack_expect_put(exp);
1348         return err;
1349 }
1350
1351 static int
1352 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1353                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1354 {
1355         struct ip_conntrack_expect *exp, *tmp;
1356         struct ip_conntrack_tuple tuple;
1357         struct ip_conntrack_helper *h;
1358         int err;
1359
1360         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1361                 return -EINVAL;
1362
1363         if (cda[CTA_EXPECT_TUPLE-1]) {
1364                 /* delete a single expect by tuple */
1365                 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1366                 if (err < 0)
1367                         return err;
1368
1369                 /* bump usage count to 2 */
1370                 exp = ip_conntrack_expect_find(&tuple);
1371                 if (!exp)
1372                         return -ENOENT;
1373
1374                 if (cda[CTA_EXPECT_ID-1]) {
1375                         u_int32_t id =
1376                                 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1377                         if (exp->id != ntohl(id)) {
1378                                 ip_conntrack_expect_put(exp);
1379                                 return -ENOENT;
1380                         }
1381                 }
1382
1383                 /* after list removal, usage count == 1 */
1384                 ip_conntrack_unexpect_related(exp);
1385                 /* have to put what we 'get' above.
1386                  * after this line usage count == 0 */
1387                 ip_conntrack_expect_put(exp);
1388         } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1389                 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1390
1391                 /* delete all expectations for this helper */
1392                 write_lock_bh(&ip_conntrack_lock);
1393                 h = __ip_conntrack_helper_find_byname(name);
1394                 if (!h) {
1395                         write_unlock_bh(&ip_conntrack_lock);
1396                         return -EINVAL;
1397                 }
1398                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1399                                          list) {
1400                         if (exp->master->helper == h
1401                             && del_timer(&exp->timeout)) {
1402                                 ip_ct_unlink_expect(exp);
1403                                 ip_conntrack_expect_put(exp);
1404                         }
1405                 }
1406                 write_unlock_bh(&ip_conntrack_lock);
1407         } else {
1408                 /* This basically means we have to flush everything*/
1409                 write_lock_bh(&ip_conntrack_lock);
1410                 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1411                                          list) {
1412                         if (del_timer(&exp->timeout)) {
1413                                 ip_ct_unlink_expect(exp);
1414                                 ip_conntrack_expect_put(exp);
1415                         }
1416                 }
1417                 write_unlock_bh(&ip_conntrack_lock);
1418         }
1419
1420         return 0;
1421 }
1422 static int
1423 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1424 {
1425         return -EOPNOTSUPP;
1426 }
1427
1428 static int
1429 ctnetlink_create_expect(struct nfattr *cda[])
1430 {
1431         struct ip_conntrack_tuple tuple, mask, master_tuple;
1432         struct ip_conntrack_tuple_hash *h = NULL;
1433         struct ip_conntrack_expect *exp;
1434         struct ip_conntrack *ct;
1435         int err = 0;
1436
1437         DEBUGP("entered %s\n", __FUNCTION__);
1438
1439         /* caller guarantees that those three CTA_EXPECT_* exist */
1440         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1441         if (err < 0)
1442                 return err;
1443         err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1444         if (err < 0)
1445                 return err;
1446         err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1447         if (err < 0)
1448                 return err;
1449
1450         /* Look for master conntrack of this expectation */
1451         h = ip_conntrack_find_get(&master_tuple, NULL);
1452         if (!h)
1453                 return -ENOENT;
1454         ct = tuplehash_to_ctrack(h);
1455
1456         if (!ct->helper) {
1457                 /* such conntrack hasn't got any helper, abort */
1458                 err = -EINVAL;
1459                 goto out;
1460         }
1461
1462         exp = ip_conntrack_expect_alloc(ct);
1463         if (!exp) {
1464                 err = -ENOMEM;
1465                 goto out;
1466         }
1467
1468         exp->expectfn = NULL;
1469         exp->flags = 0;
1470         exp->master = ct;
1471         memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1472         memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1473
1474         err = ip_conntrack_expect_related(exp);
1475         ip_conntrack_expect_put(exp);
1476
1477 out:
1478         ip_conntrack_put(tuplehash_to_ctrack(h));
1479         return err;
1480 }
1481
1482 static int
1483 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1484                      struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1485 {
1486         struct ip_conntrack_tuple tuple;
1487         struct ip_conntrack_expect *exp;
1488         int err = 0;
1489
1490         DEBUGP("entered %s\n", __FUNCTION__);
1491
1492         if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1493                 return -EINVAL;
1494
1495         if (!cda[CTA_EXPECT_TUPLE-1]
1496             || !cda[CTA_EXPECT_MASK-1]
1497             || !cda[CTA_EXPECT_MASTER-1])
1498                 return -EINVAL;
1499
1500         err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1501         if (err < 0)
1502                 return err;
1503
1504         write_lock_bh(&ip_conntrack_lock);
1505         exp = __ip_conntrack_expect_find(&tuple);
1506
1507         if (!exp) {
1508                 write_unlock_bh(&ip_conntrack_lock);
1509                 err = -ENOENT;
1510                 if (nlh->nlmsg_flags & NLM_F_CREATE)
1511                         err = ctnetlink_create_expect(cda);
1512                 return err;
1513         }
1514
1515         err = -EEXIST;
1516         if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1517                 err = ctnetlink_change_expect(exp, cda);
1518         write_unlock_bh(&ip_conntrack_lock);
1519
1520         DEBUGP("leaving\n");
1521
1522         return err;
1523 }
1524
1525 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1526 static struct notifier_block ctnl_notifier = {
1527         .notifier_call  = ctnetlink_conntrack_event,
1528 };
1529
1530 static struct notifier_block ctnl_notifier_exp = {
1531         .notifier_call  = ctnetlink_expect_event,
1532 };
1533 #endif
1534
1535 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1536         [IPCTNL_MSG_CT_NEW]             = { .call = ctnetlink_new_conntrack,
1537                                             .attr_count = CTA_MAX, },
1538         [IPCTNL_MSG_CT_GET]             = { .call = ctnetlink_get_conntrack,
1539                                             .attr_count = CTA_MAX, },
1540         [IPCTNL_MSG_CT_DELETE]          = { .call = ctnetlink_del_conntrack,
1541                                             .attr_count = CTA_MAX, },
1542         [IPCTNL_MSG_CT_GET_CTRZERO]     = { .call = ctnetlink_get_conntrack,
1543                                             .attr_count = CTA_MAX, },
1544 };
1545
1546 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1547         [IPCTNL_MSG_EXP_GET]            = { .call = ctnetlink_get_expect,
1548                                             .attr_count = CTA_EXPECT_MAX, },
1549         [IPCTNL_MSG_EXP_NEW]            = { .call = ctnetlink_new_expect,
1550                                             .attr_count = CTA_EXPECT_MAX, },
1551         [IPCTNL_MSG_EXP_DELETE]         = { .call = ctnetlink_del_expect,
1552                                             .attr_count = CTA_EXPECT_MAX, },
1553 };
1554
1555 static struct nfnetlink_subsystem ctnl_subsys = {
1556         .name                           = "conntrack",
1557         .subsys_id                      = NFNL_SUBSYS_CTNETLINK,
1558         .cb_count                       = IPCTNL_MSG_MAX,
1559         .cb                             = ctnl_cb,
1560 };
1561
1562 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1563         .name                           = "conntrack_expect",
1564         .subsys_id                      = NFNL_SUBSYS_CTNETLINK_EXP,
1565         .cb_count                       = IPCTNL_MSG_EXP_MAX,
1566         .cb                             = ctnl_exp_cb,
1567 };
1568
1569 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1570 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1571
1572 static int __init ctnetlink_init(void)
1573 {
1574         int ret;
1575
1576         printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1577         ret = nfnetlink_subsys_register(&ctnl_subsys);
1578         if (ret < 0) {
1579                 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1580                 goto err_out;
1581         }
1582
1583         ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1584         if (ret < 0) {
1585                 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1586                 goto err_unreg_subsys;
1587         }
1588
1589 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1590         ret = ip_conntrack_register_notifier(&ctnl_notifier);
1591         if (ret < 0) {
1592                 printk("ctnetlink_init: cannot register notifier.\n");
1593                 goto err_unreg_exp_subsys;
1594         }
1595
1596         ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1597         if (ret < 0) {
1598                 printk("ctnetlink_init: cannot expect register notifier.\n");
1599                 goto err_unreg_notifier;
1600         }
1601 #endif
1602
1603         return 0;
1604
1605 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1606 err_unreg_notifier:
1607         ip_conntrack_unregister_notifier(&ctnl_notifier);
1608 err_unreg_exp_subsys:
1609         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1610 #endif
1611 err_unreg_subsys:
1612         nfnetlink_subsys_unregister(&ctnl_subsys);
1613 err_out:
1614         return ret;
1615 }
1616
1617 static void __exit ctnetlink_exit(void)
1618 {
1619         printk("ctnetlink: unregistering from nfnetlink.\n");
1620
1621 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1622         ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1623         ip_conntrack_unregister_notifier(&ctnl_notifier);
1624 #endif
1625
1626         nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1627         nfnetlink_subsys_unregister(&ctnl_subsys);
1628         return;
1629 }
1630
1631 module_init(ctnetlink_init);
1632 module_exit(ctnetlink_exit);