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) 2005-2006 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/interrupt.h>
31 #include <linux/notifier.h>
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>
40 #include <linux/netfilter/nfnetlink.h>
41 #include <linux/netfilter/nfnetlink_conntrack.h>
43 MODULE_LICENSE("GPL");
45 static char __initdata version[] = "0.90";
48 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
49 const struct ip_conntrack_tuple *tuple,
50 struct ip_conntrack_protocol *proto)
53 struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
55 NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
57 if (likely(proto->tuple_to_nfattr))
58 ret = proto->tuple_to_nfattr(skb, tuple);
60 NFA_NEST_END(skb, nest_parms);
69 ctnetlink_dump_tuples_ip(struct sk_buff *skb,
70 const struct ip_conntrack_tuple *tuple)
72 struct nfattr *nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
74 NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(__be32), &tuple->src.ip);
75 NFA_PUT(skb, CTA_IP_V4_DST, sizeof(__be32), &tuple->dst.ip);
77 NFA_NEST_END(skb, nest_parms);
86 ctnetlink_dump_tuples(struct sk_buff *skb,
87 const struct ip_conntrack_tuple *tuple)
90 struct ip_conntrack_protocol *proto;
92 ret = ctnetlink_dump_tuples_ip(skb, tuple);
93 if (unlikely(ret < 0))
96 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
97 ret = ctnetlink_dump_tuples_proto(skb, tuple, proto);
98 ip_conntrack_proto_put(proto);
104 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
106 __be32 status = htonl((u_int32_t) ct->status);
107 NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
115 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
117 long timeout_l = ct->timeout.expires - jiffies;
123 timeout = htonl(timeout_l / HZ);
125 NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
133 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
135 struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
137 struct nfattr *nest_proto;
140 if (!proto->to_nfattr) {
141 ip_conntrack_proto_put(proto);
145 nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
147 ret = proto->to_nfattr(skb, nest_proto, ct);
149 ip_conntrack_proto_put(proto);
151 NFA_NEST_END(skb, nest_proto);
156 ip_conntrack_proto_put(proto);
161 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
163 struct nfattr *nest_helper;
168 nest_helper = NFA_NEST(skb, CTA_HELP);
169 NFA_PUT(skb, CTA_HELP_NAME, strlen(ct->helper->name), ct->helper->name);
171 if (ct->helper->to_nfattr)
172 ct->helper->to_nfattr(skb, ct);
174 NFA_NEST_END(skb, nest_helper);
182 #ifdef CONFIG_IP_NF_CT_ACCT
184 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
185 enum ip_conntrack_dir dir)
187 enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
188 struct nfattr *nest_count = NFA_NEST(skb, type);
191 tmp = htonl(ct->counters[dir].packets);
192 NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(__be32), &tmp);
194 tmp = htonl(ct->counters[dir].bytes);
195 NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(__be32), &tmp);
197 NFA_NEST_END(skb, nest_count);
205 #define ctnetlink_dump_counters(a, b, c) (0)
208 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
210 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
212 __be32 mark = htonl(ct->mark);
214 NFA_PUT(skb, CTA_MARK, sizeof(__be32), &mark);
221 #define ctnetlink_dump_mark(a, b) (0)
225 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
227 __be32 id = htonl(ct->id);
228 NFA_PUT(skb, CTA_ID, sizeof(__be32), &id);
236 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
238 __be32 use = htonl(atomic_read(&ct->ct_general.use));
240 NFA_PUT(skb, CTA_USE, sizeof(__be32), &use);
247 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
250 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
251 int event, int nowait,
252 const struct ip_conntrack *ct)
254 struct nlmsghdr *nlh;
255 struct nfgenmsg *nfmsg;
256 struct nfattr *nest_parms;
261 event |= NFNL_SUBSYS_CTNETLINK << 8;
262 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
263 nfmsg = NLMSG_DATA(nlh);
265 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
266 nfmsg->nfgen_family = AF_INET;
267 nfmsg->version = NFNETLINK_V0;
270 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
271 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
273 NFA_NEST_END(skb, nest_parms);
275 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
276 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
278 NFA_NEST_END(skb, nest_parms);
280 if (ctnetlink_dump_status(skb, ct) < 0 ||
281 ctnetlink_dump_timeout(skb, ct) < 0 ||
282 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
283 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
284 ctnetlink_dump_protoinfo(skb, ct) < 0 ||
285 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
286 ctnetlink_dump_mark(skb, ct) < 0 ||
287 ctnetlink_dump_id(skb, ct) < 0 ||
288 ctnetlink_dump_use(skb, ct) < 0)
291 nlh->nlmsg_len = skb->tail - b;
296 skb_trim(skb, b - skb->data);
300 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
301 static int ctnetlink_conntrack_event(struct notifier_block *this,
302 unsigned long events, void *ptr)
304 struct nlmsghdr *nlh;
305 struct nfgenmsg *nfmsg;
306 struct nfattr *nest_parms;
307 struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
311 unsigned int flags = 0, group;
313 /* ignore our fake conntrack entry */
314 if (ct == &ip_conntrack_untracked)
317 if (events & IPCT_DESTROY) {
318 type = IPCTNL_MSG_CT_DELETE;
319 group = NFNLGRP_CONNTRACK_DESTROY;
320 } else if (events & (IPCT_NEW | IPCT_RELATED)) {
321 type = IPCTNL_MSG_CT_NEW;
322 flags = NLM_F_CREATE|NLM_F_EXCL;
323 /* dump everything */
325 group = NFNLGRP_CONNTRACK_NEW;
326 } else if (events & (IPCT_STATUS | IPCT_PROTOINFO)) {
327 type = IPCTNL_MSG_CT_NEW;
328 group = NFNLGRP_CONNTRACK_UPDATE;
332 if (!nfnetlink_has_listeners(group))
335 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
341 type |= NFNL_SUBSYS_CTNETLINK << 8;
342 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
343 nfmsg = NLMSG_DATA(nlh);
345 nlh->nlmsg_flags = flags;
346 nfmsg->nfgen_family = AF_INET;
347 nfmsg->version = NFNETLINK_V0;
350 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
351 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
353 NFA_NEST_END(skb, nest_parms);
355 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
356 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
358 NFA_NEST_END(skb, nest_parms);
360 /* NAT stuff is now a status flag */
361 if ((events & IPCT_STATUS || events & IPCT_NATINFO)
362 && ctnetlink_dump_status(skb, ct) < 0)
364 if (events & IPCT_REFRESH
365 && ctnetlink_dump_timeout(skb, ct) < 0)
367 if (events & IPCT_PROTOINFO
368 && ctnetlink_dump_protoinfo(skb, ct) < 0)
370 if (events & IPCT_HELPINFO
371 && ctnetlink_dump_helpinfo(skb, ct) < 0)
374 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
375 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
378 if (events & IPCT_MARK
379 && ctnetlink_dump_mark(skb, ct) < 0)
382 nlh->nlmsg_len = skb->tail - b;
383 nfnetlink_send(skb, 0, group, 0);
391 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
393 static int ctnetlink_done(struct netlink_callback *cb)
396 ip_conntrack_put((struct ip_conntrack *)cb->args[1]);
401 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
403 struct ip_conntrack *ct, *last;
404 struct ip_conntrack_tuple_hash *h;
407 read_lock_bh(&ip_conntrack_lock);
408 last = (struct ip_conntrack *)cb->args[1];
409 for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++) {
411 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
412 h = (struct ip_conntrack_tuple_hash *) i;
413 if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
415 ct = tuplehash_to_ctrack(h);
421 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
425 nf_conntrack_get(&ct->ct_general);
426 cb->args[1] = (unsigned long)ct;
429 #ifdef CONFIG_NF_CT_ACCT
430 if (NFNL_MSG_TYPE(cb->nlh->nlmsg_type) ==
431 IPCTNL_MSG_CT_GET_CTRZERO)
432 memset(&ct->counters, 0, sizeof(ct->counters));
441 read_unlock_bh(&ip_conntrack_lock);
443 ip_conntrack_put(last);
448 static const size_t cta_min_ip[CTA_IP_MAX] = {
449 [CTA_IP_V4_SRC-1] = sizeof(__be32),
450 [CTA_IP_V4_DST-1] = sizeof(__be32),
454 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
456 struct nfattr *tb[CTA_IP_MAX];
458 nfattr_parse_nested(tb, CTA_IP_MAX, attr);
460 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
463 if (!tb[CTA_IP_V4_SRC-1])
465 tuple->src.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
467 if (!tb[CTA_IP_V4_DST-1])
469 tuple->dst.ip = *(__be32 *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
474 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
475 [CTA_PROTO_NUM-1] = sizeof(u_int8_t),
476 [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t),
477 [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t),
478 [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
479 [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
480 [CTA_PROTO_ICMP_ID-1] = sizeof(u_int16_t),
484 ctnetlink_parse_tuple_proto(struct nfattr *attr,
485 struct ip_conntrack_tuple *tuple)
487 struct nfattr *tb[CTA_PROTO_MAX];
488 struct ip_conntrack_protocol *proto;
491 nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
493 if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
496 if (!tb[CTA_PROTO_NUM-1])
498 tuple->dst.protonum = *(u_int8_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
500 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
502 if (likely(proto->nfattr_to_tuple))
503 ret = proto->nfattr_to_tuple(tb, tuple);
505 ip_conntrack_proto_put(proto);
511 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
512 enum ctattr_tuple type)
514 struct nfattr *tb[CTA_TUPLE_MAX];
517 memset(tuple, 0, sizeof(*tuple));
519 nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
521 if (!tb[CTA_TUPLE_IP-1])
524 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
528 if (!tb[CTA_TUPLE_PROTO-1])
531 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
535 /* orig and expect tuples get DIR_ORIGINAL */
536 if (type == CTA_TUPLE_REPLY)
537 tuple->dst.dir = IP_CT_DIR_REPLY;
539 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
544 #ifdef CONFIG_IP_NF_NAT_NEEDED
545 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
546 [CTA_PROTONAT_PORT_MIN-1] = sizeof(u_int16_t),
547 [CTA_PROTONAT_PORT_MAX-1] = sizeof(u_int16_t),
550 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
551 const struct ip_conntrack *ct,
552 struct ip_nat_range *range)
554 struct nfattr *tb[CTA_PROTONAT_MAX];
555 struct ip_nat_protocol *npt;
557 nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
559 if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
562 npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
564 if (!npt->nfattr_to_range) {
565 ip_nat_proto_put(npt);
569 /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
570 if (npt->nfattr_to_range(tb, range) > 0)
571 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
573 ip_nat_proto_put(npt);
578 static const size_t cta_min_nat[CTA_NAT_MAX] = {
579 [CTA_NAT_MINIP-1] = sizeof(__be32),
580 [CTA_NAT_MAXIP-1] = sizeof(__be32),
584 ctnetlink_parse_nat(struct nfattr *nat,
585 const struct ip_conntrack *ct, struct ip_nat_range *range)
587 struct nfattr *tb[CTA_NAT_MAX];
590 memset(range, 0, sizeof(*range));
592 nfattr_parse_nested(tb, CTA_NAT_MAX, nat);
594 if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
597 if (tb[CTA_NAT_MINIP-1])
598 range->min_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
600 if (!tb[CTA_NAT_MAXIP-1])
601 range->max_ip = range->min_ip;
603 range->max_ip = *(__be32 *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
606 range->flags |= IP_NAT_RANGE_MAP_IPS;
608 if (!tb[CTA_NAT_PROTO-1])
611 err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
620 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
622 struct nfattr *tb[CTA_HELP_MAX];
624 nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
626 if (!tb[CTA_HELP_NAME-1])
629 *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
634 static const size_t cta_min[CTA_MAX] = {
635 [CTA_STATUS-1] = sizeof(__be32),
636 [CTA_TIMEOUT-1] = sizeof(__be32),
637 [CTA_MARK-1] = sizeof(__be32),
638 [CTA_USE-1] = sizeof(__be32),
639 [CTA_ID-1] = sizeof(__be32)
643 ctnetlink_del_conntrack(struct sock *ctnl, struct sk_buff *skb,
644 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
646 struct ip_conntrack_tuple_hash *h;
647 struct ip_conntrack_tuple tuple;
648 struct ip_conntrack *ct;
651 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
654 if (cda[CTA_TUPLE_ORIG-1])
655 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
656 else if (cda[CTA_TUPLE_REPLY-1])
657 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
659 /* Flush the whole table */
660 ip_conntrack_flush();
667 h = ip_conntrack_find_get(&tuple, NULL);
671 ct = tuplehash_to_ctrack(h);
674 u_int32_t id = ntohl(*(__be32 *)NFA_DATA(cda[CTA_ID-1]));
676 ip_conntrack_put(ct);
680 if (del_timer(&ct->timeout))
681 ct->timeout.function((unsigned long)ct);
683 ip_conntrack_put(ct);
689 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
690 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
692 struct ip_conntrack_tuple_hash *h;
693 struct ip_conntrack_tuple tuple;
694 struct ip_conntrack *ct;
695 struct sk_buff *skb2 = NULL;
698 if (nlh->nlmsg_flags & NLM_F_DUMP) {
699 struct nfgenmsg *msg = NLMSG_DATA(nlh);
702 if (msg->nfgen_family != AF_INET)
703 return -EAFNOSUPPORT;
705 #ifndef CONFIG_IP_NF_CT_ACCT
706 if (NFNL_MSG_TYPE(nlh->nlmsg_type) == IPCTNL_MSG_CT_GET_CTRZERO)
709 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
710 ctnetlink_dump_table,
711 ctnetlink_done)) != 0)
714 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
721 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
724 if (cda[CTA_TUPLE_ORIG-1])
725 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
726 else if (cda[CTA_TUPLE_REPLY-1])
727 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
734 h = ip_conntrack_find_get(&tuple, NULL);
738 ct = tuplehash_to_ctrack(h);
741 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
743 ip_conntrack_put(ct);
747 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
748 IPCTNL_MSG_CT_NEW, 1, ct);
749 ip_conntrack_put(ct);
753 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
766 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
769 unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
770 d = ct->status ^ status;
772 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
776 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
777 /* SEEN_REPLY bit can only be set */
781 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
782 /* ASSURED bit can only be set */
785 if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
786 #ifndef CONFIG_IP_NF_NAT_NEEDED
789 struct ip_nat_range range;
791 if (cda[CTA_NAT_DST-1]) {
792 if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
795 if (ip_nat_initialized(ct,
796 HOOK2MANIP(NF_IP_PRE_ROUTING)))
798 ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
800 if (cda[CTA_NAT_SRC-1]) {
801 if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
804 if (ip_nat_initialized(ct,
805 HOOK2MANIP(NF_IP_POST_ROUTING)))
807 ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
812 /* Be careful here, modifying NAT bits can screw up things,
813 * so don't let users modify them directly if they don't pass
815 ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
821 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
823 struct ip_conntrack_helper *helper;
827 /* don't change helper of sibling connections */
831 err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
835 helper = __ip_conntrack_helper_find_byname(helpname);
837 if (!strcmp(helpname, ""))
845 /* we had a helper before ... */
846 ip_ct_remove_expectations(ct);
849 /* need to zero data of old helper */
850 memset(&ct->help, 0, sizeof(ct->help));
860 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
862 u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
864 if (!del_timer(&ct->timeout))
867 ct->timeout.expires = jiffies + timeout * HZ;
868 add_timer(&ct->timeout);
874 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
876 struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
877 struct ip_conntrack_protocol *proto;
878 u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
881 nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
883 proto = ip_conntrack_proto_find_get(npt);
885 if (proto->from_nfattr)
886 err = proto->from_nfattr(tb, ct);
887 ip_conntrack_proto_put(proto);
893 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
897 if (cda[CTA_HELP-1]) {
898 err = ctnetlink_change_helper(ct, cda);
903 if (cda[CTA_TIMEOUT-1]) {
904 err = ctnetlink_change_timeout(ct, cda);
909 if (cda[CTA_STATUS-1]) {
910 err = ctnetlink_change_status(ct, cda);
915 if (cda[CTA_PROTOINFO-1]) {
916 err = ctnetlink_change_protoinfo(ct, cda);
921 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
923 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
930 ctnetlink_create_conntrack(struct nfattr *cda[],
931 struct ip_conntrack_tuple *otuple,
932 struct ip_conntrack_tuple *rtuple)
934 struct ip_conntrack *ct;
937 ct = ip_conntrack_alloc(otuple, rtuple);
938 if (ct == NULL || IS_ERR(ct))
941 if (!cda[CTA_TIMEOUT-1])
943 ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
945 ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
946 ct->status |= IPS_CONFIRMED;
948 err = ctnetlink_change_status(ct, cda);
952 if (cda[CTA_PROTOINFO-1]) {
953 err = ctnetlink_change_protoinfo(ct, cda);
958 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
960 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
963 ct->helper = ip_conntrack_helper_find_get(rtuple);
965 add_timer(&ct->timeout);
966 ip_conntrack_hash_insert(ct);
969 ip_conntrack_helper_put(ct->helper);
974 ip_conntrack_free(ct);
979 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
980 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
982 struct ip_conntrack_tuple otuple, rtuple;
983 struct ip_conntrack_tuple_hash *h = NULL;
986 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
989 if (cda[CTA_TUPLE_ORIG-1]) {
990 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
995 if (cda[CTA_TUPLE_REPLY-1]) {
996 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1001 write_lock_bh(&ip_conntrack_lock);
1002 if (cda[CTA_TUPLE_ORIG-1])
1003 h = __ip_conntrack_find(&otuple, NULL);
1004 else if (cda[CTA_TUPLE_REPLY-1])
1005 h = __ip_conntrack_find(&rtuple, NULL);
1008 write_unlock_bh(&ip_conntrack_lock);
1010 if (nlh->nlmsg_flags & NLM_F_CREATE)
1011 err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1014 /* implicit 'else' */
1016 /* we only allow nat config for new conntracks */
1017 if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1022 /* We manipulate the conntrack inside the global conntrack table lock,
1023 * so there's no need to increase the refcount */
1025 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1026 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1029 write_unlock_bh(&ip_conntrack_lock);
1033 /***********************************************************************
1035 ***********************************************************************/
1038 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1039 const struct ip_conntrack_tuple *tuple,
1040 enum ctattr_expect type)
1042 struct nfattr *nest_parms = NFA_NEST(skb, type);
1044 if (ctnetlink_dump_tuples(skb, tuple) < 0)
1045 goto nfattr_failure;
1047 NFA_NEST_END(skb, nest_parms);
1056 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1057 const struct ip_conntrack_tuple *tuple,
1058 const struct ip_conntrack_tuple *mask)
1061 struct ip_conntrack_protocol *proto;
1062 struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1064 ret = ctnetlink_dump_tuples_ip(skb, mask);
1065 if (unlikely(ret < 0))
1066 goto nfattr_failure;
1068 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1069 ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1070 ip_conntrack_proto_put(proto);
1071 if (unlikely(ret < 0))
1072 goto nfattr_failure;
1074 NFA_NEST_END(skb, nest_parms);
1083 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1084 const struct ip_conntrack_expect *exp)
1086 struct ip_conntrack *master = exp->master;
1087 __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1088 __be32 id = htonl(exp->id);
1090 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1091 goto nfattr_failure;
1092 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1093 goto nfattr_failure;
1094 if (ctnetlink_exp_dump_tuple(skb,
1095 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1096 CTA_EXPECT_MASTER) < 0)
1097 goto nfattr_failure;
1099 NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1100 NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1109 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1112 const struct ip_conntrack_expect *exp)
1114 struct nlmsghdr *nlh;
1115 struct nfgenmsg *nfmsg;
1120 event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1121 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1122 nfmsg = NLMSG_DATA(nlh);
1124 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
1125 nfmsg->nfgen_family = AF_INET;
1126 nfmsg->version = NFNETLINK_V0;
1129 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1130 goto nfattr_failure;
1132 nlh->nlmsg_len = skb->tail - b;
1137 skb_trim(skb, b - skb->data);
1141 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1142 static int ctnetlink_expect_event(struct notifier_block *this,
1143 unsigned long events, void *ptr)
1145 struct nlmsghdr *nlh;
1146 struct nfgenmsg *nfmsg;
1147 struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1148 struct sk_buff *skb;
1153 if (events & IPEXP_NEW) {
1154 type = IPCTNL_MSG_EXP_NEW;
1155 flags = NLM_F_CREATE|NLM_F_EXCL;
1159 if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1162 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1168 type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1169 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1170 nfmsg = NLMSG_DATA(nlh);
1172 nlh->nlmsg_flags = flags;
1173 nfmsg->nfgen_family = AF_INET;
1174 nfmsg->version = NFNETLINK_V0;
1177 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1178 goto nfattr_failure;
1180 nlh->nlmsg_len = skb->tail - b;
1181 nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1192 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1194 struct ip_conntrack_expect *exp = NULL;
1195 struct list_head *i;
1196 u_int32_t *id = (u_int32_t *) &cb->args[0];
1198 read_lock_bh(&ip_conntrack_lock);
1199 list_for_each_prev(i, &ip_conntrack_expect_list) {
1200 exp = (struct ip_conntrack_expect *) i;
1203 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1211 read_unlock_bh(&ip_conntrack_lock);
1216 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1217 [CTA_EXPECT_TIMEOUT-1] = sizeof(__be32),
1218 [CTA_EXPECT_ID-1] = sizeof(__be32)
1222 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1223 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1225 struct ip_conntrack_tuple tuple;
1226 struct ip_conntrack_expect *exp;
1227 struct sk_buff *skb2;
1230 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1233 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1234 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1237 if (msg->nfgen_family != AF_INET)
1238 return -EAFNOSUPPORT;
1240 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1241 ctnetlink_exp_dump_table,
1242 ctnetlink_done)) != 0)
1244 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1245 if (rlen > skb->len)
1247 skb_pull(skb, rlen);
1251 if (cda[CTA_EXPECT_MASTER-1])
1252 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1259 exp = ip_conntrack_expect_find(&tuple);
1263 if (cda[CTA_EXPECT_ID-1]) {
1264 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1265 if (exp->id != ntohl(id)) {
1266 ip_conntrack_expect_put(exp);
1272 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1276 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1277 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1282 ip_conntrack_expect_put(exp);
1284 return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1289 ip_conntrack_expect_put(exp);
1294 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1295 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1297 struct ip_conntrack_expect *exp, *tmp;
1298 struct ip_conntrack_tuple tuple;
1299 struct ip_conntrack_helper *h;
1302 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1305 if (cda[CTA_EXPECT_TUPLE-1]) {
1306 /* delete a single expect by tuple */
1307 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1311 /* bump usage count to 2 */
1312 exp = ip_conntrack_expect_find(&tuple);
1316 if (cda[CTA_EXPECT_ID-1]) {
1318 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1319 if (exp->id != ntohl(id)) {
1320 ip_conntrack_expect_put(exp);
1325 /* after list removal, usage count == 1 */
1326 ip_conntrack_unexpect_related(exp);
1327 /* have to put what we 'get' above.
1328 * after this line usage count == 0 */
1329 ip_conntrack_expect_put(exp);
1330 } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1331 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1333 /* delete all expectations for this helper */
1334 write_lock_bh(&ip_conntrack_lock);
1335 h = __ip_conntrack_helper_find_byname(name);
1337 write_unlock_bh(&ip_conntrack_lock);
1340 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1342 if (exp->master->helper == h
1343 && del_timer(&exp->timeout)) {
1344 ip_ct_unlink_expect(exp);
1345 ip_conntrack_expect_put(exp);
1348 write_unlock_bh(&ip_conntrack_lock);
1350 /* This basically means we have to flush everything*/
1351 write_lock_bh(&ip_conntrack_lock);
1352 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1354 if (del_timer(&exp->timeout)) {
1355 ip_ct_unlink_expect(exp);
1356 ip_conntrack_expect_put(exp);
1359 write_unlock_bh(&ip_conntrack_lock);
1365 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1371 ctnetlink_create_expect(struct nfattr *cda[])
1373 struct ip_conntrack_tuple tuple, mask, master_tuple;
1374 struct ip_conntrack_tuple_hash *h = NULL;
1375 struct ip_conntrack_expect *exp;
1376 struct ip_conntrack *ct;
1379 /* caller guarantees that those three CTA_EXPECT_* exist */
1380 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1383 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1386 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1390 /* Look for master conntrack of this expectation */
1391 h = ip_conntrack_find_get(&master_tuple, NULL);
1394 ct = tuplehash_to_ctrack(h);
1397 /* such conntrack hasn't got any helper, abort */
1402 exp = ip_conntrack_expect_alloc(ct);
1408 exp->expectfn = NULL;
1411 memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1412 memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1414 err = ip_conntrack_expect_related(exp);
1415 ip_conntrack_expect_put(exp);
1418 ip_conntrack_put(tuplehash_to_ctrack(h));
1423 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1424 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1426 struct ip_conntrack_tuple tuple;
1427 struct ip_conntrack_expect *exp;
1430 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1433 if (!cda[CTA_EXPECT_TUPLE-1]
1434 || !cda[CTA_EXPECT_MASK-1]
1435 || !cda[CTA_EXPECT_MASTER-1])
1438 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1442 write_lock_bh(&ip_conntrack_lock);
1443 exp = __ip_conntrack_expect_find(&tuple);
1446 write_unlock_bh(&ip_conntrack_lock);
1448 if (nlh->nlmsg_flags & NLM_F_CREATE)
1449 err = ctnetlink_create_expect(cda);
1454 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1455 err = ctnetlink_change_expect(exp, cda);
1456 write_unlock_bh(&ip_conntrack_lock);
1461 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1462 static struct notifier_block ctnl_notifier = {
1463 .notifier_call = ctnetlink_conntrack_event,
1466 static struct notifier_block ctnl_notifier_exp = {
1467 .notifier_call = ctnetlink_expect_event,
1471 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1472 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
1473 .attr_count = CTA_MAX, },
1474 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
1475 .attr_count = CTA_MAX, },
1476 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
1477 .attr_count = CTA_MAX, },
1478 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
1479 .attr_count = CTA_MAX, },
1482 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1483 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
1484 .attr_count = CTA_EXPECT_MAX, },
1485 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
1486 .attr_count = CTA_EXPECT_MAX, },
1487 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
1488 .attr_count = CTA_EXPECT_MAX, },
1491 static struct nfnetlink_subsystem ctnl_subsys = {
1492 .name = "conntrack",
1493 .subsys_id = NFNL_SUBSYS_CTNETLINK,
1494 .cb_count = IPCTNL_MSG_MAX,
1498 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1499 .name = "conntrack_expect",
1500 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
1501 .cb_count = IPCTNL_MSG_EXP_MAX,
1505 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1506 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1508 static int __init ctnetlink_init(void)
1512 printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1513 ret = nfnetlink_subsys_register(&ctnl_subsys);
1515 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1519 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1521 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1522 goto err_unreg_subsys;
1525 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1526 ret = ip_conntrack_register_notifier(&ctnl_notifier);
1528 printk("ctnetlink_init: cannot register notifier.\n");
1529 goto err_unreg_exp_subsys;
1532 ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1534 printk("ctnetlink_init: cannot expect register notifier.\n");
1535 goto err_unreg_notifier;
1541 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1543 ip_conntrack_unregister_notifier(&ctnl_notifier);
1544 err_unreg_exp_subsys:
1545 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1548 nfnetlink_subsys_unregister(&ctnl_subsys);
1553 static void __exit ctnetlink_exit(void)
1555 printk("ctnetlink: unregistering from nfnetlink.\n");
1557 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1558 ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1559 ip_conntrack_unregister_notifier(&ctnl_notifier);
1562 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1563 nfnetlink_subsys_unregister(&ctnl_subsys);
1567 module_init(ctnetlink_init);
1568 module_exit(ctnetlink_exit);