1 /* Connection tracking via netlink socket. Allows for user space
2 * protocol helpers and general trouble making from userspace.
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>
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.
12 * Initial connection tracking via netlink development funded and
13 * generally made possible by Network Robots, Inc. (www.networkrobots.com)
15 * Further development of this code funded by Astaro AG (http://www.astaro.com)
17 * This software may be used and distributed according to the terms
18 * of the GNU General Public License, incorporated herein by reference.
21 #include <linux/init.h>
22 #include <linux/module.h>
23 #include <linux/kernel.h>
24 #include <linux/types.h>
25 #include <linux/timer.h>
26 #include <linux/skbuff.h>
27 #include <linux/errno.h>
28 #include <linux/netlink.h>
29 #include <linux/spinlock.h>
30 #include <linux/notifier.h>
31 #include <linux/rtnetlink.h>
33 #include <linux/netfilter.h>
34 #include <linux/netfilter_ipv4.h>
35 #include <linux/netfilter_ipv4/ip_tables.h>
36 #include <linux/netfilter_ipv4/ip_conntrack.h>
37 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
38 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
39 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
40 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
42 #include <linux/netfilter/nfnetlink.h>
43 #include <linux/netfilter/nfnetlink_conntrack.h>
45 MODULE_LICENSE("GPL");
47 static char __initdata version[] = "0.90";
52 #define DEBUGP(format, args...)
57 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
58 const struct ip_conntrack_tuple *tuple)
60 struct ip_conntrack_protocol *proto;
62 NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
64 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
65 if (proto && proto->tuple_to_nfattr)
66 return proto->tuple_to_nfattr(skb, tuple);
75 ctnetlink_dump_tuples(struct sk_buff *skb,
76 const struct ip_conntrack_tuple *tuple)
78 struct nfattr *nest_parms;
80 nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
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 NFA_NEST_END(skb, nest_parms);
85 nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
86 ctnetlink_dump_tuples_proto(skb, tuple);
87 NFA_NEST_END(skb, nest_parms);
96 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
98 u_int32_t status = htonl((u_int32_t) ct->status);
99 NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
107 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
109 long timeout_l = ct->timeout.expires - jiffies;
115 timeout = htonl(timeout_l / HZ);
117 NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
125 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
127 struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
129 struct nfattr *nest_proto;
132 if (!proto || !proto->to_nfattr)
135 nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
137 ret = proto->to_nfattr(skb, nest_proto, ct);
139 ip_conntrack_proto_put(proto);
141 NFA_NEST_END(skb, nest_proto);
150 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
152 struct nfattr *nest_helper;
157 nest_helper = NFA_NEST(skb, CTA_HELP);
158 NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
160 if (ct->helper->to_nfattr)
161 ct->helper->to_nfattr(skb, ct);
163 NFA_NEST_END(skb, nest_helper);
171 #ifdef CONFIG_IP_NF_CT_ACCT
173 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
174 enum ip_conntrack_dir dir)
176 enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
177 struct nfattr *nest_count = NFA_NEST(skb, type);
180 tmp = htonl(ct->counters[dir].packets);
181 NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
183 tmp = htonl(ct->counters[dir].bytes);
184 NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
186 NFA_NEST_END(skb, nest_count);
194 #define ctnetlink_dump_counters(a, b, c) (0)
197 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
199 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
201 u_int32_t mark = htonl(ct->mark);
203 NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
210 #define ctnetlink_dump_mark(a, b) (0)
214 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
216 u_int32_t id = htonl(ct->id);
217 NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
225 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
227 unsigned int use = htonl(atomic_read(&ct->ct_general.use));
229 NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
236 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
239 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
240 int event, int nowait,
241 const struct ip_conntrack *ct)
243 struct nlmsghdr *nlh;
244 struct nfgenmsg *nfmsg;
245 struct nfattr *nest_parms;
250 event |= NFNL_SUBSYS_CTNETLINK << 8;
251 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
252 nfmsg = NLMSG_DATA(nlh);
254 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
255 nfmsg->nfgen_family = AF_INET;
256 nfmsg->version = NFNETLINK_V0;
259 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
260 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
262 NFA_NEST_END(skb, nest_parms);
264 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
265 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
267 NFA_NEST_END(skb, nest_parms);
269 if (ctnetlink_dump_status(skb, ct) < 0 ||
270 ctnetlink_dump_timeout(skb, ct) < 0 ||
271 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
272 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
273 ctnetlink_dump_protoinfo(skb, ct) < 0 ||
274 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
275 ctnetlink_dump_mark(skb, ct) < 0 ||
276 ctnetlink_dump_id(skb, ct) < 0 ||
277 ctnetlink_dump_use(skb, ct) < 0)
280 nlh->nlmsg_len = skb->tail - b;
285 skb_trim(skb, b - skb->data);
289 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
290 static int ctnetlink_conntrack_event(struct notifier_block *this,
291 unsigned long events, void *ptr)
293 struct nlmsghdr *nlh;
294 struct nfgenmsg *nfmsg;
295 struct nfattr *nest_parms;
296 struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
300 unsigned int flags = 0, group;
302 /* ignore our fake conntrack entry */
303 if (ct == &ip_conntrack_untracked)
306 if (events & IPCT_DESTROY) {
307 type = IPCTNL_MSG_CT_DELETE;
308 group = NFNLGRP_CONNTRACK_DESTROY;
311 if (events & (IPCT_NEW | IPCT_RELATED)) {
312 type = IPCTNL_MSG_CT_NEW;
313 flags = NLM_F_CREATE|NLM_F_EXCL;
314 /* dump everything */
316 group = NFNLGRP_CONNTRACK_NEW;
319 if (events & (IPCT_STATUS |
324 type = IPCTNL_MSG_CT_NEW;
325 group = NFNLGRP_CONNTRACK_UPDATE;
332 /* FIXME: Check if there are any listeners before, don't hurt performance */
334 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
340 type |= NFNL_SUBSYS_CTNETLINK << 8;
341 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
342 nfmsg = NLMSG_DATA(nlh);
344 nlh->nlmsg_flags = flags;
345 nfmsg->nfgen_family = AF_INET;
346 nfmsg->version = NFNETLINK_V0;
349 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
350 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
352 NFA_NEST_END(skb, nest_parms);
354 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
355 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
357 NFA_NEST_END(skb, nest_parms);
359 /* NAT stuff is now a status flag */
360 if ((events & IPCT_STATUS || events & IPCT_NATINFO)
361 && ctnetlink_dump_status(skb, ct) < 0)
363 if (events & IPCT_REFRESH
364 && ctnetlink_dump_timeout(skb, ct) < 0)
366 if (events & IPCT_PROTOINFO
367 && ctnetlink_dump_protoinfo(skb, ct) < 0)
369 if (events & IPCT_HELPINFO
370 && ctnetlink_dump_helpinfo(skb, ct) < 0)
373 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
374 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
377 nlh->nlmsg_len = skb->tail - b;
378 nfnetlink_send(skb, 0, group, 0);
386 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
388 static int ctnetlink_done(struct netlink_callback *cb)
390 DEBUGP("entered %s\n", __FUNCTION__);
395 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
397 struct ip_conntrack *ct = NULL;
398 struct ip_conntrack_tuple_hash *h;
400 u_int32_t *id = (u_int32_t *) &cb->args[1];
402 DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
405 read_lock_bh(&ip_conntrack_lock);
406 for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
407 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
408 h = (struct ip_conntrack_tuple_hash *) i;
409 if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
411 ct = tuplehash_to_ctrack(h);
414 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
423 read_unlock_bh(&ip_conntrack_lock);
425 DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
430 #ifdef CONFIG_IP_NF_CT_ACCT
432 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
434 struct ip_conntrack *ct = NULL;
435 struct ip_conntrack_tuple_hash *h;
437 u_int32_t *id = (u_int32_t *) &cb->args[1];
439 DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
442 write_lock_bh(&ip_conntrack_lock);
443 for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
444 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
445 h = (struct ip_conntrack_tuple_hash *) i;
446 if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
448 ct = tuplehash_to_ctrack(h);
451 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
458 memset(&ct->counters, 0, sizeof(ct->counters));
462 write_unlock_bh(&ip_conntrack_lock);
464 DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
470 static const int cta_min_ip[CTA_IP_MAX] = {
471 [CTA_IP_V4_SRC-1] = sizeof(u_int32_t),
472 [CTA_IP_V4_DST-1] = sizeof(u_int32_t),
476 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
478 struct nfattr *tb[CTA_IP_MAX];
480 DEBUGP("entered %s\n", __FUNCTION__);
483 if (nfattr_parse_nested(tb, CTA_IP_MAX, attr) < 0)
486 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
489 if (!tb[CTA_IP_V4_SRC-1])
491 tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
493 if (!tb[CTA_IP_V4_DST-1])
495 tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
505 static const int cta_min_proto[CTA_PROTO_MAX] = {
506 [CTA_PROTO_NUM-1] = sizeof(u_int16_t),
507 [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t),
508 [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t),
509 [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
510 [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
511 [CTA_PROTO_ICMP_ID-1] = sizeof(u_int16_t),
515 ctnetlink_parse_tuple_proto(struct nfattr *attr,
516 struct ip_conntrack_tuple *tuple)
518 struct nfattr *tb[CTA_PROTO_MAX];
519 struct ip_conntrack_protocol *proto;
522 DEBUGP("entered %s\n", __FUNCTION__);
524 if (nfattr_parse_nested(tb, CTA_PROTO_MAX, attr) < 0)
527 if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
530 if (!tb[CTA_PROTO_NUM-1])
532 tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
534 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
536 if (likely(proto && proto->nfattr_to_tuple)) {
537 ret = proto->nfattr_to_tuple(tb, tuple);
538 ip_conntrack_proto_put(proto);
548 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
549 enum ctattr_tuple type)
551 struct nfattr *tb[CTA_TUPLE_MAX];
554 DEBUGP("entered %s\n", __FUNCTION__);
556 memset(tuple, 0, sizeof(*tuple));
558 if (nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]) < 0)
561 if (!tb[CTA_TUPLE_IP-1])
564 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
568 if (!tb[CTA_TUPLE_PROTO-1])
571 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
575 /* orig and expect tuples get DIR_ORIGINAL */
576 if (type == CTA_TUPLE_REPLY)
577 tuple->dst.dir = IP_CT_DIR_REPLY;
579 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
591 #ifdef CONFIG_IP_NF_NAT_NEEDED
592 static const int cta_min_protonat[CTA_PROTONAT_MAX] = {
593 [CTA_PROTONAT_PORT_MIN-1] = sizeof(u_int16_t),
594 [CTA_PROTONAT_PORT_MAX-1] = sizeof(u_int16_t),
597 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
598 const struct ip_conntrack *ct,
599 struct ip_nat_range *range)
601 struct nfattr *tb[CTA_PROTONAT_MAX];
602 struct ip_nat_protocol *npt;
604 DEBUGP("entered %s\n", __FUNCTION__);
606 if (nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr) < 0)
609 if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
612 npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
616 if (!npt->nfattr_to_range) {
617 ip_nat_proto_put(npt);
621 /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
622 if (npt->nfattr_to_range(tb, range) > 0)
623 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
625 ip_nat_proto_put(npt);
635 ctnetlink_parse_nat(struct nfattr *cda[],
636 const struct ip_conntrack *ct, struct ip_nat_range *range)
638 struct nfattr *tb[CTA_NAT_MAX];
641 DEBUGP("entered %s\n", __FUNCTION__);
643 memset(range, 0, sizeof(*range));
645 if (nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]) < 0)
648 if (tb[CTA_NAT_MINIP-1])
649 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
651 if (!tb[CTA_NAT_MAXIP-1])
652 range->max_ip = range->min_ip;
654 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
657 range->flags |= IP_NAT_RANGE_MAP_IPS;
659 if (!tb[CTA_NAT_PROTO-1])
662 err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
675 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
677 struct nfattr *tb[CTA_HELP_MAX];
679 DEBUGP("entered %s\n", __FUNCTION__);
681 if (nfattr_parse_nested(tb, CTA_HELP_MAX, attr) < 0)
684 if (!tb[CTA_HELP_NAME-1])
687 *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
696 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
697 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
699 struct ip_conntrack_tuple_hash *h;
700 struct ip_conntrack_tuple tuple;
701 struct ip_conntrack *ct;
704 DEBUGP("entered %s\n", __FUNCTION__);
706 if (cda[CTA_TUPLE_ORIG-1])
707 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
708 else if (cda[CTA_TUPLE_REPLY-1])
709 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
711 /* Flush the whole table */
712 ip_conntrack_flush();
719 h = ip_conntrack_find_get(&tuple, NULL);
721 DEBUGP("tuple not found in conntrack hash\n");
725 ct = tuplehash_to_ctrack(h);
728 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
730 ip_conntrack_put(ct);
734 if (del_timer(&ct->timeout)) {
735 ip_conntrack_put(ct);
736 ct->timeout.function((unsigned long)ct);
739 ip_conntrack_put(ct);
746 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
747 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
749 struct ip_conntrack_tuple_hash *h;
750 struct ip_conntrack_tuple tuple;
751 struct ip_conntrack *ct;
752 struct sk_buff *skb2 = NULL;
755 DEBUGP("entered %s\n", __FUNCTION__);
757 if (nlh->nlmsg_flags & NLM_F_DUMP) {
758 struct nfgenmsg *msg = NLMSG_DATA(nlh);
761 if (msg->nfgen_family != AF_INET)
762 return -EAFNOSUPPORT;
764 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
765 IPCTNL_MSG_CT_GET_CTRZERO) {
766 #ifdef CONFIG_IP_NF_CT_ACCT
767 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
768 ctnetlink_dump_table_w,
769 ctnetlink_done)) != 0)
775 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
776 ctnetlink_dump_table,
777 ctnetlink_done)) != 0)
781 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
788 if (cda[CTA_TUPLE_ORIG-1])
789 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
790 else if (cda[CTA_TUPLE_REPLY-1])
791 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
798 h = ip_conntrack_find_get(&tuple, NULL);
800 DEBUGP("tuple not found in conntrack hash");
803 DEBUGP("tuple found\n");
804 ct = tuplehash_to_ctrack(h);
807 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
809 ip_conntrack_put(ct);
812 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
814 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
815 IPCTNL_MSG_CT_NEW, 1, ct);
816 ip_conntrack_put(ct);
820 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
834 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
837 unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
838 d = ct->status ^ status;
840 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
844 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
845 /* SEEN_REPLY bit can only be set */
849 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
850 /* ASSURED bit can only be set */
853 if (cda[CTA_NAT-1]) {
854 #ifndef CONFIG_IP_NF_NAT_NEEDED
857 unsigned int hooknum;
858 struct ip_nat_range range;
860 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
863 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
864 NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
865 htons(range.min.all), htons(range.max.all));
867 /* This is tricky but it works. ip_nat_setup_info needs the
868 * hook number as parameter, so let's do the correct
869 * conversion and run away */
870 if (status & IPS_SRC_NAT_DONE)
871 hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
872 else if (status & IPS_DST_NAT_DONE)
873 hooknum = NF_IP_PRE_ROUTING; /* IP_NAT_MANIP_DST */
875 return -EINVAL; /* Missing NAT flags */
877 DEBUGP("NAT status: %lu\n",
878 status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
880 if (ip_nat_initialized(ct, hooknum))
882 ip_nat_setup_info(ct, &range, hooknum);
884 DEBUGP("NAT status after setup_info: %lu\n",
885 ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
889 /* Be careful here, modifying NAT bits can screw up things,
890 * so don't let users modify them directly if they don't pass
892 ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
898 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
900 struct ip_conntrack_helper *helper;
904 DEBUGP("entered %s\n", __FUNCTION__);
906 /* don't change helper of sibling connections */
910 err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
914 helper = __ip_conntrack_helper_find_byname(helpname);
916 if (!strcmp(helpname, ""))
924 /* we had a helper before ... */
925 ip_ct_remove_expectations(ct);
928 /* need to zero data of old helper */
929 memset(&ct->help, 0, sizeof(ct->help));
939 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
941 u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
943 if (!del_timer(&ct->timeout))
946 ct->timeout.expires = jiffies + timeout * HZ;
947 add_timer(&ct->timeout);
953 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
955 struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
956 struct ip_conntrack_protocol *proto;
957 u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
960 if (nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr) < 0)
963 proto = ip_conntrack_proto_find_get(npt);
967 if (proto->from_nfattr)
968 err = proto->from_nfattr(tb, ct);
969 ip_conntrack_proto_put(proto);
978 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
982 DEBUGP("entered %s\n", __FUNCTION__);
984 if (cda[CTA_HELP-1]) {
985 err = ctnetlink_change_helper(ct, cda);
990 if (cda[CTA_TIMEOUT-1]) {
991 err = ctnetlink_change_timeout(ct, cda);
996 if (cda[CTA_STATUS-1]) {
997 err = ctnetlink_change_status(ct, cda);
1002 if (cda[CTA_PROTOINFO-1]) {
1003 err = ctnetlink_change_protoinfo(ct, cda);
1008 DEBUGP("all done\n");
1013 ctnetlink_create_conntrack(struct nfattr *cda[],
1014 struct ip_conntrack_tuple *otuple,
1015 struct ip_conntrack_tuple *rtuple)
1017 struct ip_conntrack *ct;
1020 DEBUGP("entered %s\n", __FUNCTION__);
1022 ct = ip_conntrack_alloc(otuple, rtuple);
1023 if (ct == NULL || IS_ERR(ct))
1026 if (!cda[CTA_TIMEOUT-1])
1028 ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1030 ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1031 ct->status |= IPS_CONFIRMED;
1033 err = ctnetlink_change_status(ct, cda);
1037 if (cda[CTA_PROTOINFO-1]) {
1038 err = ctnetlink_change_protoinfo(ct, cda);
1043 ct->helper = ip_conntrack_helper_find_get(rtuple);
1045 add_timer(&ct->timeout);
1046 ip_conntrack_hash_insert(ct);
1049 ip_conntrack_helper_put(ct->helper);
1051 DEBUGP("conntrack with id %u inserted\n", ct->id);
1055 ip_conntrack_free(ct);
1060 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1061 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1063 struct ip_conntrack_tuple otuple, rtuple;
1064 struct ip_conntrack_tuple_hash *h = NULL;
1067 DEBUGP("entered %s\n", __FUNCTION__);
1069 if (cda[CTA_TUPLE_ORIG-1]) {
1070 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1075 if (cda[CTA_TUPLE_REPLY-1]) {
1076 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1081 write_lock_bh(&ip_conntrack_lock);
1082 if (cda[CTA_TUPLE_ORIG-1])
1083 h = __ip_conntrack_find(&otuple, NULL);
1084 else if (cda[CTA_TUPLE_REPLY-1])
1085 h = __ip_conntrack_find(&rtuple, NULL);
1088 write_unlock_bh(&ip_conntrack_lock);
1089 DEBUGP("no such conntrack, create new\n");
1091 if (nlh->nlmsg_flags & NLM_F_CREATE)
1092 err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1095 /* implicit 'else' */
1097 /* we only allow nat config for new conntracks */
1098 if (cda[CTA_NAT-1]) {
1103 /* We manipulate the conntrack inside the global conntrack table lock,
1104 * so there's no need to increase the refcount */
1105 DEBUGP("conntrack found\n");
1107 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1108 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1111 write_unlock_bh(&ip_conntrack_lock);
1115 /***********************************************************************
1117 ***********************************************************************/
1120 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1121 const struct ip_conntrack_tuple *tuple,
1122 enum ctattr_expect type)
1124 struct nfattr *nest_parms = NFA_NEST(skb, type);
1126 if (ctnetlink_dump_tuples(skb, tuple) < 0)
1127 goto nfattr_failure;
1129 NFA_NEST_END(skb, nest_parms);
1138 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1139 const struct ip_conntrack_expect *exp)
1141 struct ip_conntrack *master = exp->master;
1142 u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1143 u_int32_t id = htonl(exp->id);
1145 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1146 goto nfattr_failure;
1147 if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1148 goto nfattr_failure;
1149 if (ctnetlink_exp_dump_tuple(skb,
1150 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1151 CTA_EXPECT_MASTER) < 0)
1152 goto nfattr_failure;
1154 NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1155 NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1164 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1167 const struct ip_conntrack_expect *exp)
1169 struct nlmsghdr *nlh;
1170 struct nfgenmsg *nfmsg;
1175 event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1176 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1177 nfmsg = NLMSG_DATA(nlh);
1179 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
1180 nfmsg->nfgen_family = AF_INET;
1181 nfmsg->version = NFNETLINK_V0;
1184 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1185 goto nfattr_failure;
1187 nlh->nlmsg_len = skb->tail - b;
1192 skb_trim(skb, b - skb->data);
1196 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1197 static int ctnetlink_expect_event(struct notifier_block *this,
1198 unsigned long events, void *ptr)
1200 struct nlmsghdr *nlh;
1201 struct nfgenmsg *nfmsg;
1202 struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1203 struct sk_buff *skb;
1209 if (events & IPEXP_NEW) {
1210 type = IPCTNL_MSG_EXP_NEW;
1211 flags = NLM_F_CREATE|NLM_F_EXCL;
1215 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1221 type |= NFNL_SUBSYS_CTNETLINK << 8;
1222 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1223 nfmsg = NLMSG_DATA(nlh);
1225 nlh->nlmsg_flags = flags;
1226 nfmsg->nfgen_family = AF_INET;
1227 nfmsg->version = NFNETLINK_V0;
1230 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1231 goto nfattr_failure;
1233 nlh->nlmsg_len = skb->tail - b;
1234 proto = exp->tuple.dst.protonum;
1235 nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1246 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1248 struct ip_conntrack_expect *exp = NULL;
1249 struct list_head *i;
1250 u_int32_t *id = (u_int32_t *) &cb->args[0];
1252 DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1254 read_lock_bh(&ip_conntrack_lock);
1255 list_for_each_prev(i, &ip_conntrack_expect_list) {
1256 exp = (struct ip_conntrack_expect *) i;
1259 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1267 read_unlock_bh(&ip_conntrack_lock);
1269 DEBUGP("leaving, last id=%llu\n", *id);
1275 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1276 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1278 struct ip_conntrack_tuple tuple;
1279 struct ip_conntrack_expect *exp;
1280 struct sk_buff *skb2;
1283 DEBUGP("entered %s\n", __FUNCTION__);
1285 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1286 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1289 if (msg->nfgen_family != AF_INET)
1290 return -EAFNOSUPPORT;
1292 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1293 ctnetlink_exp_dump_table,
1294 ctnetlink_done)) != 0)
1296 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1297 if (rlen > skb->len)
1299 skb_pull(skb, rlen);
1303 if (cda[CTA_EXPECT_MASTER-1])
1304 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1311 exp = ip_conntrack_expect_find(&tuple);
1316 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1319 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1321 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1322 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1327 ip_conntrack_expect_put(exp);
1329 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1336 ip_conntrack_expect_put(exp);
1344 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1345 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1347 struct ip_conntrack_expect *exp, *tmp;
1348 struct ip_conntrack_tuple tuple;
1349 struct ip_conntrack_helper *h;
1352 if (cda[CTA_EXPECT_TUPLE-1]) {
1353 /* delete a single expect by tuple */
1354 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1358 /* bump usage count to 2 */
1359 exp = ip_conntrack_expect_find(&tuple);
1363 if (cda[CTA_EXPECT_ID-1]) {
1365 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1366 if (exp->id != ntohl(id)) {
1367 ip_conntrack_expect_put(exp);
1372 /* after list removal, usage count == 1 */
1373 ip_conntrack_unexpect_related(exp);
1374 /* have to put what we 'get' above.
1375 * after this line usage count == 0 */
1376 ip_conntrack_expect_put(exp);
1377 } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1378 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1380 /* delete all expectations for this helper */
1381 write_lock_bh(&ip_conntrack_lock);
1382 h = __ip_conntrack_helper_find_byname(name);
1384 write_unlock_bh(&ip_conntrack_lock);
1387 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1389 if (exp->master->helper == h
1390 && del_timer(&exp->timeout)) {
1391 ip_ct_unlink_expect(exp);
1392 ip_conntrack_expect_put(exp);
1395 write_unlock(&ip_conntrack_lock);
1397 /* This basically means we have to flush everything*/
1398 write_lock_bh(&ip_conntrack_lock);
1399 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1401 if (del_timer(&exp->timeout)) {
1402 ip_ct_unlink_expect(exp);
1403 ip_conntrack_expect_put(exp);
1406 write_unlock_bh(&ip_conntrack_lock);
1412 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1418 ctnetlink_create_expect(struct nfattr *cda[])
1420 struct ip_conntrack_tuple tuple, mask, master_tuple;
1421 struct ip_conntrack_tuple_hash *h = NULL;
1422 struct ip_conntrack_expect *exp;
1423 struct ip_conntrack *ct;
1426 DEBUGP("entered %s\n", __FUNCTION__);
1428 /* caller guarantees that those three CTA_EXPECT_* exist */
1429 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1432 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1435 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1439 /* Look for master conntrack of this expectation */
1440 h = ip_conntrack_find_get(&master_tuple, NULL);
1443 ct = tuplehash_to_ctrack(h);
1446 /* such conntrack hasn't got any helper, abort */
1451 exp = ip_conntrack_expect_alloc(ct);
1457 exp->expectfn = NULL;
1460 memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1461 memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1463 err = ip_conntrack_expect_related(exp);
1464 ip_conntrack_expect_put(exp);
1467 ip_conntrack_put(tuplehash_to_ctrack(h));
1472 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1473 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1475 struct ip_conntrack_tuple tuple;
1476 struct ip_conntrack_expect *exp;
1479 DEBUGP("entered %s\n", __FUNCTION__);
1481 if (!cda[CTA_EXPECT_TUPLE-1]
1482 || !cda[CTA_EXPECT_MASK-1]
1483 || !cda[CTA_EXPECT_MASTER-1])
1486 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1490 write_lock_bh(&ip_conntrack_lock);
1491 exp = __ip_conntrack_expect_find(&tuple);
1494 write_unlock_bh(&ip_conntrack_lock);
1496 if (nlh->nlmsg_flags & NLM_F_CREATE)
1497 err = ctnetlink_create_expect(cda);
1502 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1503 err = ctnetlink_change_expect(exp, cda);
1504 write_unlock_bh(&ip_conntrack_lock);
1506 DEBUGP("leaving\n");
1511 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1512 static struct notifier_block ctnl_notifier = {
1513 .notifier_call = ctnetlink_conntrack_event,
1516 static struct notifier_block ctnl_notifier_exp = {
1517 .notifier_call = ctnetlink_expect_event,
1521 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1522 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
1523 .attr_count = CTA_MAX,
1524 .cap_required = CAP_NET_ADMIN },
1525 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
1526 .attr_count = CTA_MAX,
1527 .cap_required = CAP_NET_ADMIN },
1528 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
1529 .attr_count = CTA_MAX,
1530 .cap_required = CAP_NET_ADMIN },
1531 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
1532 .attr_count = CTA_MAX,
1533 .cap_required = CAP_NET_ADMIN },
1536 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1537 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
1538 .attr_count = CTA_EXPECT_MAX,
1539 .cap_required = CAP_NET_ADMIN },
1540 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
1541 .attr_count = CTA_EXPECT_MAX,
1542 .cap_required = CAP_NET_ADMIN },
1543 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
1544 .attr_count = CTA_EXPECT_MAX,
1545 .cap_required = CAP_NET_ADMIN },
1548 static struct nfnetlink_subsystem ctnl_subsys = {
1549 .name = "conntrack",
1550 .subsys_id = NFNL_SUBSYS_CTNETLINK,
1551 .cb_count = IPCTNL_MSG_MAX,
1555 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1556 .name = "conntrack_expect",
1557 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
1558 .cb_count = IPCTNL_MSG_EXP_MAX,
1562 static int __init ctnetlink_init(void)
1566 printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1567 ret = nfnetlink_subsys_register(&ctnl_subsys);
1569 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1573 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1575 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1576 goto err_unreg_subsys;
1579 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1580 ret = ip_conntrack_register_notifier(&ctnl_notifier);
1582 printk("ctnetlink_init: cannot register notifier.\n");
1583 goto err_unreg_exp_subsys;
1586 ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1588 printk("ctnetlink_init: cannot expect register notifier.\n");
1589 goto err_unreg_notifier;
1595 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1597 ip_conntrack_unregister_notifier(&ctnl_notifier);
1598 err_unreg_exp_subsys:
1599 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1602 nfnetlink_subsys_unregister(&ctnl_subsys);
1607 static void __exit ctnetlink_exit(void)
1609 printk("ctnetlink: unregistering from nfnetlink.\n");
1611 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1612 ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1613 ip_conntrack_unregister_notifier(&ctnl_notifier);
1616 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1617 nfnetlink_subsys_unregister(&ctnl_subsys);
1621 module_init(ctnetlink_init);
1622 module_exit(ctnetlink_exit);