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);
746 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
748 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
749 IPCTNL_MSG_CT_NEW, 1, ct);
750 ip_conntrack_put(ct);
754 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
767 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
770 unsigned status = ntohl(*(__be32 *)NFA_DATA(cda[CTA_STATUS-1]));
771 d = ct->status ^ status;
773 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
777 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
778 /* SEEN_REPLY bit can only be set */
782 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
783 /* ASSURED bit can only be set */
786 if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
787 #ifndef CONFIG_IP_NF_NAT_NEEDED
790 struct ip_nat_range range;
792 if (cda[CTA_NAT_DST-1]) {
793 if (ctnetlink_parse_nat(cda[CTA_NAT_DST-1], ct,
796 if (ip_nat_initialized(ct,
797 HOOK2MANIP(NF_IP_PRE_ROUTING)))
799 ip_nat_setup_info(ct, &range, NF_IP_PRE_ROUTING);
801 if (cda[CTA_NAT_SRC-1]) {
802 if (ctnetlink_parse_nat(cda[CTA_NAT_SRC-1], ct,
805 if (ip_nat_initialized(ct,
806 HOOK2MANIP(NF_IP_POST_ROUTING)))
808 ip_nat_setup_info(ct, &range, NF_IP_POST_ROUTING);
813 /* Be careful here, modifying NAT bits can screw up things,
814 * so don't let users modify them directly if they don't pass
816 ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
822 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
824 struct ip_conntrack_helper *helper;
828 /* don't change helper of sibling connections */
832 err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
836 helper = __ip_conntrack_helper_find_byname(helpname);
838 if (!strcmp(helpname, ""))
846 /* we had a helper before ... */
847 ip_ct_remove_expectations(ct);
850 /* need to zero data of old helper */
851 memset(&ct->help, 0, sizeof(ct->help));
861 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
863 u_int32_t timeout = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
865 if (!del_timer(&ct->timeout))
868 ct->timeout.expires = jiffies + timeout * HZ;
869 add_timer(&ct->timeout);
875 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
877 struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
878 struct ip_conntrack_protocol *proto;
879 u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
882 nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
884 proto = ip_conntrack_proto_find_get(npt);
886 if (proto->from_nfattr)
887 err = proto->from_nfattr(tb, ct);
888 ip_conntrack_proto_put(proto);
894 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
898 if (cda[CTA_HELP-1]) {
899 err = ctnetlink_change_helper(ct, cda);
904 if (cda[CTA_TIMEOUT-1]) {
905 err = ctnetlink_change_timeout(ct, cda);
910 if (cda[CTA_STATUS-1]) {
911 err = ctnetlink_change_status(ct, cda);
916 if (cda[CTA_PROTOINFO-1]) {
917 err = ctnetlink_change_protoinfo(ct, cda);
922 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
924 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
931 ctnetlink_create_conntrack(struct nfattr *cda[],
932 struct ip_conntrack_tuple *otuple,
933 struct ip_conntrack_tuple *rtuple)
935 struct ip_conntrack *ct;
938 ct = ip_conntrack_alloc(otuple, rtuple);
939 if (ct == NULL || IS_ERR(ct))
942 if (!cda[CTA_TIMEOUT-1])
944 ct->timeout.expires = ntohl(*(__be32 *)NFA_DATA(cda[CTA_TIMEOUT-1]));
946 ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
947 ct->status |= IPS_CONFIRMED;
949 err = ctnetlink_change_status(ct, cda);
953 if (cda[CTA_PROTOINFO-1]) {
954 err = ctnetlink_change_protoinfo(ct, cda);
959 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
961 ct->mark = ntohl(*(__be32 *)NFA_DATA(cda[CTA_MARK-1]));
964 ct->helper = ip_conntrack_helper_find_get(rtuple);
966 add_timer(&ct->timeout);
967 ip_conntrack_hash_insert(ct);
970 ip_conntrack_helper_put(ct->helper);
975 ip_conntrack_free(ct);
980 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
981 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
983 struct ip_conntrack_tuple otuple, rtuple;
984 struct ip_conntrack_tuple_hash *h = NULL;
987 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
990 if (cda[CTA_TUPLE_ORIG-1]) {
991 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
996 if (cda[CTA_TUPLE_REPLY-1]) {
997 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1002 write_lock_bh(&ip_conntrack_lock);
1003 if (cda[CTA_TUPLE_ORIG-1])
1004 h = __ip_conntrack_find(&otuple, NULL);
1005 else if (cda[CTA_TUPLE_REPLY-1])
1006 h = __ip_conntrack_find(&rtuple, NULL);
1009 write_unlock_bh(&ip_conntrack_lock);
1011 if (nlh->nlmsg_flags & NLM_F_CREATE)
1012 err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1015 /* implicit 'else' */
1017 /* we only allow nat config for new conntracks */
1018 if (cda[CTA_NAT_SRC-1] || cda[CTA_NAT_DST-1]) {
1023 /* We manipulate the conntrack inside the global conntrack table lock,
1024 * so there's no need to increase the refcount */
1026 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1027 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1030 write_unlock_bh(&ip_conntrack_lock);
1034 /***********************************************************************
1036 ***********************************************************************/
1039 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1040 const struct ip_conntrack_tuple *tuple,
1041 enum ctattr_expect type)
1043 struct nfattr *nest_parms = NFA_NEST(skb, type);
1045 if (ctnetlink_dump_tuples(skb, tuple) < 0)
1046 goto nfattr_failure;
1048 NFA_NEST_END(skb, nest_parms);
1057 ctnetlink_exp_dump_mask(struct sk_buff *skb,
1058 const struct ip_conntrack_tuple *tuple,
1059 const struct ip_conntrack_tuple *mask)
1062 struct ip_conntrack_protocol *proto;
1063 struct nfattr *nest_parms = NFA_NEST(skb, CTA_EXPECT_MASK);
1065 ret = ctnetlink_dump_tuples_ip(skb, mask);
1066 if (unlikely(ret < 0))
1067 goto nfattr_failure;
1069 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
1070 ret = ctnetlink_dump_tuples_proto(skb, mask, proto);
1071 ip_conntrack_proto_put(proto);
1072 if (unlikely(ret < 0))
1073 goto nfattr_failure;
1075 NFA_NEST_END(skb, nest_parms);
1084 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1085 const struct ip_conntrack_expect *exp)
1087 struct ip_conntrack *master = exp->master;
1088 __be32 timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1089 __be32 id = htonl(exp->id);
1091 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1092 goto nfattr_failure;
1093 if (ctnetlink_exp_dump_mask(skb, &exp->tuple, &exp->mask) < 0)
1094 goto nfattr_failure;
1095 if (ctnetlink_exp_dump_tuple(skb,
1096 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1097 CTA_EXPECT_MASTER) < 0)
1098 goto nfattr_failure;
1100 NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(__be32), &timeout);
1101 NFA_PUT(skb, CTA_EXPECT_ID, sizeof(__be32), &id);
1110 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1113 const struct ip_conntrack_expect *exp)
1115 struct nlmsghdr *nlh;
1116 struct nfgenmsg *nfmsg;
1121 event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1122 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1123 nfmsg = NLMSG_DATA(nlh);
1125 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
1126 nfmsg->nfgen_family = AF_INET;
1127 nfmsg->version = NFNETLINK_V0;
1130 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1131 goto nfattr_failure;
1133 nlh->nlmsg_len = skb->tail - b;
1138 skb_trim(skb, b - skb->data);
1142 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1143 static int ctnetlink_expect_event(struct notifier_block *this,
1144 unsigned long events, void *ptr)
1146 struct nlmsghdr *nlh;
1147 struct nfgenmsg *nfmsg;
1148 struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1149 struct sk_buff *skb;
1154 if (events & IPEXP_NEW) {
1155 type = IPCTNL_MSG_EXP_NEW;
1156 flags = NLM_F_CREATE|NLM_F_EXCL;
1160 if (!nfnetlink_has_listeners(NFNLGRP_CONNTRACK_EXP_NEW))
1163 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1169 type |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1170 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1171 nfmsg = NLMSG_DATA(nlh);
1173 nlh->nlmsg_flags = flags;
1174 nfmsg->nfgen_family = AF_INET;
1175 nfmsg->version = NFNETLINK_V0;
1178 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1179 goto nfattr_failure;
1181 nlh->nlmsg_len = skb->tail - b;
1182 nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1193 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1195 struct ip_conntrack_expect *exp = NULL;
1196 struct list_head *i;
1197 u_int32_t *id = (u_int32_t *) &cb->args[0];
1199 read_lock_bh(&ip_conntrack_lock);
1200 list_for_each_prev(i, &ip_conntrack_expect_list) {
1201 exp = (struct ip_conntrack_expect *) i;
1204 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1212 read_unlock_bh(&ip_conntrack_lock);
1217 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1218 [CTA_EXPECT_TIMEOUT-1] = sizeof(__be32),
1219 [CTA_EXPECT_ID-1] = sizeof(__be32)
1223 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1224 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1226 struct ip_conntrack_tuple tuple;
1227 struct ip_conntrack_expect *exp;
1228 struct sk_buff *skb2;
1231 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1234 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1235 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1238 if (msg->nfgen_family != AF_INET)
1239 return -EAFNOSUPPORT;
1241 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1242 ctnetlink_exp_dump_table,
1243 ctnetlink_done)) != 0)
1245 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1246 if (rlen > skb->len)
1248 skb_pull(skb, rlen);
1252 if (cda[CTA_EXPECT_MASTER-1])
1253 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1260 exp = ip_conntrack_expect_find(&tuple);
1264 if (cda[CTA_EXPECT_ID-1]) {
1265 __be32 id = *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1266 if (exp->id != ntohl(id)) {
1267 ip_conntrack_expect_put(exp);
1273 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1276 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1278 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1279 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1284 ip_conntrack_expect_put(exp);
1286 return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1291 ip_conntrack_expect_put(exp);
1296 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1297 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1299 struct ip_conntrack_expect *exp, *tmp;
1300 struct ip_conntrack_tuple tuple;
1301 struct ip_conntrack_helper *h;
1304 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1307 if (cda[CTA_EXPECT_TUPLE-1]) {
1308 /* delete a single expect by tuple */
1309 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1313 /* bump usage count to 2 */
1314 exp = ip_conntrack_expect_find(&tuple);
1318 if (cda[CTA_EXPECT_ID-1]) {
1320 *(__be32 *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1321 if (exp->id != ntohl(id)) {
1322 ip_conntrack_expect_put(exp);
1327 /* after list removal, usage count == 1 */
1328 ip_conntrack_unexpect_related(exp);
1329 /* have to put what we 'get' above.
1330 * after this line usage count == 0 */
1331 ip_conntrack_expect_put(exp);
1332 } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1333 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1335 /* delete all expectations for this helper */
1336 write_lock_bh(&ip_conntrack_lock);
1337 h = __ip_conntrack_helper_find_byname(name);
1339 write_unlock_bh(&ip_conntrack_lock);
1342 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1344 if (exp->master->helper == h
1345 && del_timer(&exp->timeout)) {
1346 ip_ct_unlink_expect(exp);
1347 ip_conntrack_expect_put(exp);
1350 write_unlock_bh(&ip_conntrack_lock);
1352 /* This basically means we have to flush everything*/
1353 write_lock_bh(&ip_conntrack_lock);
1354 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1356 if (del_timer(&exp->timeout)) {
1357 ip_ct_unlink_expect(exp);
1358 ip_conntrack_expect_put(exp);
1361 write_unlock_bh(&ip_conntrack_lock);
1367 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1373 ctnetlink_create_expect(struct nfattr *cda[])
1375 struct ip_conntrack_tuple tuple, mask, master_tuple;
1376 struct ip_conntrack_tuple_hash *h = NULL;
1377 struct ip_conntrack_expect *exp;
1378 struct ip_conntrack *ct;
1381 /* caller guarantees that those three CTA_EXPECT_* exist */
1382 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1385 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1388 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1392 /* Look for master conntrack of this expectation */
1393 h = ip_conntrack_find_get(&master_tuple, NULL);
1396 ct = tuplehash_to_ctrack(h);
1399 /* such conntrack hasn't got any helper, abort */
1404 exp = ip_conntrack_expect_alloc(ct);
1410 exp->expectfn = NULL;
1413 memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1414 memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1416 err = ip_conntrack_expect_related(exp);
1417 ip_conntrack_expect_put(exp);
1420 ip_conntrack_put(tuplehash_to_ctrack(h));
1425 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1426 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1428 struct ip_conntrack_tuple tuple;
1429 struct ip_conntrack_expect *exp;
1432 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1435 if (!cda[CTA_EXPECT_TUPLE-1]
1436 || !cda[CTA_EXPECT_MASK-1]
1437 || !cda[CTA_EXPECT_MASTER-1])
1440 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1444 write_lock_bh(&ip_conntrack_lock);
1445 exp = __ip_conntrack_expect_find(&tuple);
1448 write_unlock_bh(&ip_conntrack_lock);
1450 if (nlh->nlmsg_flags & NLM_F_CREATE)
1451 err = ctnetlink_create_expect(cda);
1456 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1457 err = ctnetlink_change_expect(exp, cda);
1458 write_unlock_bh(&ip_conntrack_lock);
1463 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1464 static struct notifier_block ctnl_notifier = {
1465 .notifier_call = ctnetlink_conntrack_event,
1468 static struct notifier_block ctnl_notifier_exp = {
1469 .notifier_call = ctnetlink_expect_event,
1473 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1474 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
1475 .attr_count = CTA_MAX, },
1476 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
1477 .attr_count = CTA_MAX, },
1478 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
1479 .attr_count = CTA_MAX, },
1480 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
1481 .attr_count = CTA_MAX, },
1484 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1485 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
1486 .attr_count = CTA_EXPECT_MAX, },
1487 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
1488 .attr_count = CTA_EXPECT_MAX, },
1489 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
1490 .attr_count = CTA_EXPECT_MAX, },
1493 static struct nfnetlink_subsystem ctnl_subsys = {
1494 .name = "conntrack",
1495 .subsys_id = NFNL_SUBSYS_CTNETLINK,
1496 .cb_count = IPCTNL_MSG_MAX,
1500 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1501 .name = "conntrack_expect",
1502 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
1503 .cb_count = IPCTNL_MSG_EXP_MAX,
1507 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1508 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK_EXP);
1510 static int __init ctnetlink_init(void)
1514 printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1515 ret = nfnetlink_subsys_register(&ctnl_subsys);
1517 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1521 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1523 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1524 goto err_unreg_subsys;
1527 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1528 ret = ip_conntrack_register_notifier(&ctnl_notifier);
1530 printk("ctnetlink_init: cannot register notifier.\n");
1531 goto err_unreg_exp_subsys;
1534 ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1536 printk("ctnetlink_init: cannot expect register notifier.\n");
1537 goto err_unreg_notifier;
1543 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1545 ip_conntrack_unregister_notifier(&ctnl_notifier);
1546 err_unreg_exp_subsys:
1547 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1550 nfnetlink_subsys_unregister(&ctnl_subsys);
1555 static void __exit ctnetlink_exit(void)
1557 printk("ctnetlink: unregistering from nfnetlink.\n");
1559 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1560 ip_conntrack_expect_unregister_notifier(&ctnl_notifier_exp);
1561 ip_conntrack_unregister_notifier(&ctnl_notifier);
1564 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1565 nfnetlink_subsys_unregister(&ctnl_subsys);
1569 module_init(ctnetlink_init);
1570 module_exit(ctnetlink_exit);