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>
32 #include <linux/netfilter.h>
33 #include <linux/netfilter_ipv4/ip_conntrack.h>
34 #include <linux/netfilter_ipv4/ip_conntrack_core.h>
35 #include <linux/netfilter_ipv4/ip_conntrack_helper.h>
36 #include <linux/netfilter_ipv4/ip_conntrack_protocol.h>
37 #include <linux/netfilter_ipv4/ip_nat_protocol.h>
39 #include <linux/netfilter/nfnetlink.h>
40 #include <linux/netfilter/nfnetlink_conntrack.h>
42 MODULE_LICENSE("GPL");
44 static char __initdata version[] = "0.90";
49 #define DEBUGP(format, args...)
54 ctnetlink_dump_tuples_proto(struct sk_buff *skb,
55 const struct ip_conntrack_tuple *tuple)
57 struct ip_conntrack_protocol *proto;
60 NFA_PUT(skb, CTA_PROTO_NUM, sizeof(u_int8_t), &tuple->dst.protonum);
62 /* If no protocol helper is found, this function will return the
63 * generic protocol helper, so proto won't *ever* be NULL */
64 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
65 if (likely(proto->tuple_to_nfattr))
66 ret = proto->tuple_to_nfattr(skb, tuple);
68 ip_conntrack_proto_put(proto);
77 ctnetlink_dump_tuples(struct sk_buff *skb,
78 const struct ip_conntrack_tuple *tuple)
80 struct nfattr *nest_parms;
82 nest_parms = NFA_NEST(skb, CTA_TUPLE_IP);
83 NFA_PUT(skb, CTA_IP_V4_SRC, sizeof(u_int32_t), &tuple->src.ip);
84 NFA_PUT(skb, CTA_IP_V4_DST, sizeof(u_int32_t), &tuple->dst.ip);
85 NFA_NEST_END(skb, nest_parms);
87 nest_parms = NFA_NEST(skb, CTA_TUPLE_PROTO);
88 ctnetlink_dump_tuples_proto(skb, tuple);
89 NFA_NEST_END(skb, nest_parms);
98 ctnetlink_dump_status(struct sk_buff *skb, const struct ip_conntrack *ct)
100 u_int32_t status = htonl((u_int32_t) ct->status);
101 NFA_PUT(skb, CTA_STATUS, sizeof(status), &status);
109 ctnetlink_dump_timeout(struct sk_buff *skb, const struct ip_conntrack *ct)
111 long timeout_l = ct->timeout.expires - jiffies;
117 timeout = htonl(timeout_l / HZ);
119 NFA_PUT(skb, CTA_TIMEOUT, sizeof(timeout), &timeout);
127 ctnetlink_dump_protoinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
129 struct ip_conntrack_protocol *proto = ip_conntrack_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
131 struct nfattr *nest_proto;
134 if (!proto->to_nfattr) {
135 ip_conntrack_proto_put(proto);
139 nest_proto = NFA_NEST(skb, CTA_PROTOINFO);
141 ret = proto->to_nfattr(skb, nest_proto, ct);
143 ip_conntrack_proto_put(proto);
145 NFA_NEST_END(skb, nest_proto);
154 ctnetlink_dump_helpinfo(struct sk_buff *skb, const struct ip_conntrack *ct)
156 struct nfattr *nest_helper;
161 nest_helper = NFA_NEST(skb, CTA_HELP);
162 NFA_PUT(skb, CTA_HELP_NAME, CTA_HELP_MAXNAMESIZE, &ct->helper->name);
164 if (ct->helper->to_nfattr)
165 ct->helper->to_nfattr(skb, ct);
167 NFA_NEST_END(skb, nest_helper);
175 #ifdef CONFIG_IP_NF_CT_ACCT
177 ctnetlink_dump_counters(struct sk_buff *skb, const struct ip_conntrack *ct,
178 enum ip_conntrack_dir dir)
180 enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
181 struct nfattr *nest_count = NFA_NEST(skb, type);
184 tmp = htonl(ct->counters[dir].packets);
185 NFA_PUT(skb, CTA_COUNTERS32_PACKETS, sizeof(u_int32_t), &tmp);
187 tmp = htonl(ct->counters[dir].bytes);
188 NFA_PUT(skb, CTA_COUNTERS32_BYTES, sizeof(u_int32_t), &tmp);
190 NFA_NEST_END(skb, nest_count);
198 #define ctnetlink_dump_counters(a, b, c) (0)
201 #ifdef CONFIG_IP_NF_CONNTRACK_MARK
203 ctnetlink_dump_mark(struct sk_buff *skb, const struct ip_conntrack *ct)
205 u_int32_t mark = htonl(ct->mark);
207 NFA_PUT(skb, CTA_MARK, sizeof(u_int32_t), &mark);
214 #define ctnetlink_dump_mark(a, b) (0)
218 ctnetlink_dump_id(struct sk_buff *skb, const struct ip_conntrack *ct)
220 u_int32_t id = htonl(ct->id);
221 NFA_PUT(skb, CTA_ID, sizeof(u_int32_t), &id);
229 ctnetlink_dump_use(struct sk_buff *skb, const struct ip_conntrack *ct)
231 unsigned int use = htonl(atomic_read(&ct->ct_general.use));
233 NFA_PUT(skb, CTA_USE, sizeof(u_int32_t), &use);
240 #define tuple(ct, dir) (&(ct)->tuplehash[dir].tuple)
243 ctnetlink_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
244 int event, int nowait,
245 const struct ip_conntrack *ct)
247 struct nlmsghdr *nlh;
248 struct nfgenmsg *nfmsg;
249 struct nfattr *nest_parms;
254 event |= NFNL_SUBSYS_CTNETLINK << 8;
255 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
256 nfmsg = NLMSG_DATA(nlh);
258 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
259 nfmsg->nfgen_family = AF_INET;
260 nfmsg->version = NFNETLINK_V0;
263 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
264 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
266 NFA_NEST_END(skb, nest_parms);
268 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
269 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
271 NFA_NEST_END(skb, nest_parms);
273 if (ctnetlink_dump_status(skb, ct) < 0 ||
274 ctnetlink_dump_timeout(skb, ct) < 0 ||
275 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
276 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0 ||
277 ctnetlink_dump_protoinfo(skb, ct) < 0 ||
278 ctnetlink_dump_helpinfo(skb, ct) < 0 ||
279 ctnetlink_dump_mark(skb, ct) < 0 ||
280 ctnetlink_dump_id(skb, ct) < 0 ||
281 ctnetlink_dump_use(skb, ct) < 0)
284 nlh->nlmsg_len = skb->tail - b;
289 skb_trim(skb, b - skb->data);
293 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
294 static int ctnetlink_conntrack_event(struct notifier_block *this,
295 unsigned long events, void *ptr)
297 struct nlmsghdr *nlh;
298 struct nfgenmsg *nfmsg;
299 struct nfattr *nest_parms;
300 struct ip_conntrack *ct = (struct ip_conntrack *)ptr;
304 unsigned int flags = 0, group;
306 /* ignore our fake conntrack entry */
307 if (ct == &ip_conntrack_untracked)
310 if (events & IPCT_DESTROY) {
311 type = IPCTNL_MSG_CT_DELETE;
312 group = NFNLGRP_CONNTRACK_DESTROY;
315 if (events & (IPCT_NEW | IPCT_RELATED)) {
316 type = IPCTNL_MSG_CT_NEW;
317 flags = NLM_F_CREATE|NLM_F_EXCL;
318 /* dump everything */
320 group = NFNLGRP_CONNTRACK_NEW;
323 if (events & (IPCT_STATUS |
328 type = IPCTNL_MSG_CT_NEW;
329 group = NFNLGRP_CONNTRACK_UPDATE;
336 /* FIXME: Check if there are any listeners before, don't hurt performance */
338 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
344 type |= NFNL_SUBSYS_CTNETLINK << 8;
345 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
346 nfmsg = NLMSG_DATA(nlh);
348 nlh->nlmsg_flags = flags;
349 nfmsg->nfgen_family = AF_INET;
350 nfmsg->version = NFNETLINK_V0;
353 nest_parms = NFA_NEST(skb, CTA_TUPLE_ORIG);
354 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_ORIGINAL)) < 0)
356 NFA_NEST_END(skb, nest_parms);
358 nest_parms = NFA_NEST(skb, CTA_TUPLE_REPLY);
359 if (ctnetlink_dump_tuples(skb, tuple(ct, IP_CT_DIR_REPLY)) < 0)
361 NFA_NEST_END(skb, nest_parms);
363 /* NAT stuff is now a status flag */
364 if ((events & IPCT_STATUS || events & IPCT_NATINFO)
365 && ctnetlink_dump_status(skb, ct) < 0)
367 if (events & IPCT_REFRESH
368 && ctnetlink_dump_timeout(skb, ct) < 0)
370 if (events & IPCT_PROTOINFO
371 && ctnetlink_dump_protoinfo(skb, ct) < 0)
373 if (events & IPCT_HELPINFO
374 && ctnetlink_dump_helpinfo(skb, ct) < 0)
377 if (ctnetlink_dump_counters(skb, ct, IP_CT_DIR_ORIGINAL) < 0 ||
378 ctnetlink_dump_counters(skb, ct, IP_CT_DIR_REPLY) < 0)
381 nlh->nlmsg_len = skb->tail - b;
382 nfnetlink_send(skb, 0, group, 0);
390 #endif /* CONFIG_IP_NF_CONNTRACK_EVENTS */
392 static int ctnetlink_done(struct netlink_callback *cb)
394 DEBUGP("entered %s\n", __FUNCTION__);
399 ctnetlink_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
401 struct ip_conntrack *ct = NULL;
402 struct ip_conntrack_tuple_hash *h;
404 u_int32_t *id = (u_int32_t *) &cb->args[1];
406 DEBUGP("entered %s, last bucket=%lu id=%u\n", __FUNCTION__,
409 read_lock_bh(&ip_conntrack_lock);
410 for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 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);
418 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
427 read_unlock_bh(&ip_conntrack_lock);
429 DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
434 #ifdef CONFIG_IP_NF_CT_ACCT
436 ctnetlink_dump_table_w(struct sk_buff *skb, struct netlink_callback *cb)
438 struct ip_conntrack *ct = NULL;
439 struct ip_conntrack_tuple_hash *h;
441 u_int32_t *id = (u_int32_t *) &cb->args[1];
443 DEBUGP("entered %s, last bucket=%u id=%u\n", __FUNCTION__,
446 write_lock_bh(&ip_conntrack_lock);
447 for (; cb->args[0] < ip_conntrack_htable_size; cb->args[0]++, *id = 0) {
448 list_for_each_prev(i, &ip_conntrack_hash[cb->args[0]]) {
449 h = (struct ip_conntrack_tuple_hash *) i;
450 if (DIRECTION(h) != IP_CT_DIR_ORIGINAL)
452 ct = tuplehash_to_ctrack(h);
455 if (ctnetlink_fill_info(skb, NETLINK_CB(cb->skb).pid,
462 memset(&ct->counters, 0, sizeof(ct->counters));
466 write_unlock_bh(&ip_conntrack_lock);
468 DEBUGP("leaving, last bucket=%lu id=%u\n", cb->args[0], *id);
474 static const size_t cta_min_ip[CTA_IP_MAX] = {
475 [CTA_IP_V4_SRC-1] = sizeof(u_int32_t),
476 [CTA_IP_V4_DST-1] = sizeof(u_int32_t),
480 ctnetlink_parse_tuple_ip(struct nfattr *attr, struct ip_conntrack_tuple *tuple)
482 struct nfattr *tb[CTA_IP_MAX];
484 DEBUGP("entered %s\n", __FUNCTION__);
486 nfattr_parse_nested(tb, CTA_IP_MAX, attr);
488 if (nfattr_bad_size(tb, CTA_IP_MAX, cta_min_ip))
491 if (!tb[CTA_IP_V4_SRC-1])
493 tuple->src.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_SRC-1]);
495 if (!tb[CTA_IP_V4_DST-1])
497 tuple->dst.ip = *(u_int32_t *)NFA_DATA(tb[CTA_IP_V4_DST-1]);
504 static const size_t cta_min_proto[CTA_PROTO_MAX] = {
505 [CTA_PROTO_NUM-1] = sizeof(u_int16_t),
506 [CTA_PROTO_SRC_PORT-1] = sizeof(u_int16_t),
507 [CTA_PROTO_DST_PORT-1] = sizeof(u_int16_t),
508 [CTA_PROTO_ICMP_TYPE-1] = sizeof(u_int8_t),
509 [CTA_PROTO_ICMP_CODE-1] = sizeof(u_int8_t),
510 [CTA_PROTO_ICMP_ID-1] = sizeof(u_int16_t),
514 ctnetlink_parse_tuple_proto(struct nfattr *attr,
515 struct ip_conntrack_tuple *tuple)
517 struct nfattr *tb[CTA_PROTO_MAX];
518 struct ip_conntrack_protocol *proto;
521 DEBUGP("entered %s\n", __FUNCTION__);
523 nfattr_parse_nested(tb, CTA_PROTO_MAX, attr);
525 if (nfattr_bad_size(tb, CTA_PROTO_MAX, cta_min_proto))
528 if (!tb[CTA_PROTO_NUM-1])
530 tuple->dst.protonum = *(u_int16_t *)NFA_DATA(tb[CTA_PROTO_NUM-1]);
532 proto = ip_conntrack_proto_find_get(tuple->dst.protonum);
534 if (likely(proto->nfattr_to_tuple))
535 ret = proto->nfattr_to_tuple(tb, tuple);
537 ip_conntrack_proto_put(proto);
543 ctnetlink_parse_tuple(struct nfattr *cda[], struct ip_conntrack_tuple *tuple,
544 enum ctattr_tuple type)
546 struct nfattr *tb[CTA_TUPLE_MAX];
549 DEBUGP("entered %s\n", __FUNCTION__);
551 memset(tuple, 0, sizeof(*tuple));
553 nfattr_parse_nested(tb, CTA_TUPLE_MAX, cda[type-1]);
555 if (!tb[CTA_TUPLE_IP-1])
558 err = ctnetlink_parse_tuple_ip(tb[CTA_TUPLE_IP-1], tuple);
562 if (!tb[CTA_TUPLE_PROTO-1])
565 err = ctnetlink_parse_tuple_proto(tb[CTA_TUPLE_PROTO-1], tuple);
569 /* orig and expect tuples get DIR_ORIGINAL */
570 if (type == CTA_TUPLE_REPLY)
571 tuple->dst.dir = IP_CT_DIR_REPLY;
573 tuple->dst.dir = IP_CT_DIR_ORIGINAL;
582 #ifdef CONFIG_IP_NF_NAT_NEEDED
583 static const size_t cta_min_protonat[CTA_PROTONAT_MAX] = {
584 [CTA_PROTONAT_PORT_MIN-1] = sizeof(u_int16_t),
585 [CTA_PROTONAT_PORT_MAX-1] = sizeof(u_int16_t),
588 static int ctnetlink_parse_nat_proto(struct nfattr *attr,
589 const struct ip_conntrack *ct,
590 struct ip_nat_range *range)
592 struct nfattr *tb[CTA_PROTONAT_MAX];
593 struct ip_nat_protocol *npt;
595 DEBUGP("entered %s\n", __FUNCTION__);
597 nfattr_parse_nested(tb, CTA_PROTONAT_MAX, attr);
599 if (nfattr_bad_size(tb, CTA_PROTONAT_MAX, cta_min_protonat))
602 npt = ip_nat_proto_find_get(ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum);
604 if (!npt->nfattr_to_range) {
605 ip_nat_proto_put(npt);
609 /* nfattr_to_range returns 1 if it parsed, 0 if not, neg. on error */
610 if (npt->nfattr_to_range(tb, range) > 0)
611 range->flags |= IP_NAT_RANGE_PROTO_SPECIFIED;
613 ip_nat_proto_put(npt);
619 static const size_t cta_min_nat[CTA_NAT_MAX] = {
620 [CTA_NAT_MINIP-1] = sizeof(u_int32_t),
621 [CTA_NAT_MAXIP-1] = sizeof(u_int32_t),
625 ctnetlink_parse_nat(struct nfattr *cda[],
626 const struct ip_conntrack *ct, struct ip_nat_range *range)
628 struct nfattr *tb[CTA_NAT_MAX];
631 DEBUGP("entered %s\n", __FUNCTION__);
633 memset(range, 0, sizeof(*range));
635 nfattr_parse_nested(tb, CTA_NAT_MAX, cda[CTA_NAT-1]);
637 if (nfattr_bad_size(tb, CTA_NAT_MAX, cta_min_nat))
640 if (tb[CTA_NAT_MINIP-1])
641 range->min_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MINIP-1]);
643 if (!tb[CTA_NAT_MAXIP-1])
644 range->max_ip = range->min_ip;
646 range->max_ip = *(u_int32_t *)NFA_DATA(tb[CTA_NAT_MAXIP-1]);
649 range->flags |= IP_NAT_RANGE_MAP_IPS;
651 if (!tb[CTA_NAT_PROTO-1])
654 err = ctnetlink_parse_nat_proto(tb[CTA_NAT_PROTO-1], ct, range);
664 ctnetlink_parse_help(struct nfattr *attr, char **helper_name)
666 struct nfattr *tb[CTA_HELP_MAX];
668 DEBUGP("entered %s\n", __FUNCTION__);
670 nfattr_parse_nested(tb, CTA_HELP_MAX, attr);
672 if (!tb[CTA_HELP_NAME-1])
675 *helper_name = NFA_DATA(tb[CTA_HELP_NAME-1]);
680 static const size_t cta_min[CTA_MAX] = {
681 [CTA_STATUS-1] = sizeof(u_int32_t),
682 [CTA_TIMEOUT-1] = sizeof(u_int32_t),
683 [CTA_MARK-1] = sizeof(u_int32_t),
684 [CTA_USE-1] = sizeof(u_int32_t),
685 [CTA_ID-1] = sizeof(u_int32_t)
689 ctnetlink_del_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;
697 DEBUGP("entered %s\n", __FUNCTION__);
699 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
702 if (cda[CTA_TUPLE_ORIG-1])
703 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
704 else if (cda[CTA_TUPLE_REPLY-1])
705 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
707 /* Flush the whole table */
708 ip_conntrack_flush();
715 h = ip_conntrack_find_get(&tuple, NULL);
717 DEBUGP("tuple not found in conntrack hash\n");
721 ct = tuplehash_to_ctrack(h);
724 u_int32_t id = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_ID-1]));
726 ip_conntrack_put(ct);
730 if (del_timer(&ct->timeout)) {
731 ip_conntrack_put(ct);
732 ct->timeout.function((unsigned long)ct);
735 ip_conntrack_put(ct);
742 ctnetlink_get_conntrack(struct sock *ctnl, struct sk_buff *skb,
743 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
745 struct ip_conntrack_tuple_hash *h;
746 struct ip_conntrack_tuple tuple;
747 struct ip_conntrack *ct;
748 struct sk_buff *skb2 = NULL;
751 DEBUGP("entered %s\n", __FUNCTION__);
753 if (nlh->nlmsg_flags & NLM_F_DUMP) {
754 struct nfgenmsg *msg = NLMSG_DATA(nlh);
757 if (msg->nfgen_family != AF_INET)
758 return -EAFNOSUPPORT;
760 if (NFNL_MSG_TYPE(nlh->nlmsg_type) ==
761 IPCTNL_MSG_CT_GET_CTRZERO) {
762 #ifdef CONFIG_IP_NF_CT_ACCT
763 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
764 ctnetlink_dump_table_w,
765 ctnetlink_done)) != 0)
771 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
772 ctnetlink_dump_table,
773 ctnetlink_done)) != 0)
777 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
784 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
787 if (cda[CTA_TUPLE_ORIG-1])
788 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_ORIG);
789 else if (cda[CTA_TUPLE_REPLY-1])
790 err = ctnetlink_parse_tuple(cda, &tuple, CTA_TUPLE_REPLY);
797 h = ip_conntrack_find_get(&tuple, NULL);
799 DEBUGP("tuple not found in conntrack hash");
802 DEBUGP("tuple found\n");
803 ct = tuplehash_to_ctrack(h);
806 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
808 ip_conntrack_put(ct);
811 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
813 err = ctnetlink_fill_info(skb2, NETLINK_CB(skb).pid, nlh->nlmsg_seq,
814 IPCTNL_MSG_CT_NEW, 1, ct);
815 ip_conntrack_put(ct);
819 err = netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
833 ctnetlink_change_status(struct ip_conntrack *ct, struct nfattr *cda[])
836 unsigned status = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_STATUS-1]));
837 d = ct->status ^ status;
839 if (d & (IPS_EXPECTED|IPS_CONFIRMED|IPS_DYING))
843 if (d & IPS_SEEN_REPLY && !(status & IPS_SEEN_REPLY))
844 /* SEEN_REPLY bit can only be set */
848 if (d & IPS_ASSURED && !(status & IPS_ASSURED))
849 /* ASSURED bit can only be set */
852 if (cda[CTA_NAT-1]) {
853 #ifndef CONFIG_IP_NF_NAT_NEEDED
856 unsigned int hooknum;
857 struct ip_nat_range range;
859 if (ctnetlink_parse_nat(cda, ct, &range) < 0)
862 DEBUGP("NAT: %u.%u.%u.%u-%u.%u.%u.%u:%u-%u\n",
863 NIPQUAD(range.min_ip), NIPQUAD(range.max_ip),
864 htons(range.min.all), htons(range.max.all));
866 /* This is tricky but it works. ip_nat_setup_info needs the
867 * hook number as parameter, so let's do the correct
868 * conversion and run away */
869 if (status & IPS_SRC_NAT_DONE)
870 hooknum = NF_IP_POST_ROUTING; /* IP_NAT_MANIP_SRC */
871 else if (status & IPS_DST_NAT_DONE)
872 hooknum = NF_IP_PRE_ROUTING; /* IP_NAT_MANIP_DST */
874 return -EINVAL; /* Missing NAT flags */
876 DEBUGP("NAT status: %lu\n",
877 status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
879 if (ip_nat_initialized(ct, hooknum))
881 ip_nat_setup_info(ct, &range, hooknum);
883 DEBUGP("NAT status after setup_info: %lu\n",
884 ct->status & (IPS_NAT_MASK | IPS_NAT_DONE_MASK));
888 /* Be careful here, modifying NAT bits can screw up things,
889 * so don't let users modify them directly if they don't pass
891 ct->status |= status & ~(IPS_NAT_DONE_MASK | IPS_NAT_MASK);
897 ctnetlink_change_helper(struct ip_conntrack *ct, struct nfattr *cda[])
899 struct ip_conntrack_helper *helper;
903 DEBUGP("entered %s\n", __FUNCTION__);
905 /* don't change helper of sibling connections */
909 err = ctnetlink_parse_help(cda[CTA_HELP-1], &helpname);
913 helper = __ip_conntrack_helper_find_byname(helpname);
915 if (!strcmp(helpname, ""))
923 /* we had a helper before ... */
924 ip_ct_remove_expectations(ct);
927 /* need to zero data of old helper */
928 memset(&ct->help, 0, sizeof(ct->help));
938 ctnetlink_change_timeout(struct ip_conntrack *ct, struct nfattr *cda[])
940 u_int32_t timeout = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
942 if (!del_timer(&ct->timeout))
945 ct->timeout.expires = jiffies + timeout * HZ;
946 add_timer(&ct->timeout);
952 ctnetlink_change_protoinfo(struct ip_conntrack *ct, struct nfattr *cda[])
954 struct nfattr *tb[CTA_PROTOINFO_MAX], *attr = cda[CTA_PROTOINFO-1];
955 struct ip_conntrack_protocol *proto;
956 u_int16_t npt = ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple.dst.protonum;
959 nfattr_parse_nested(tb, CTA_PROTOINFO_MAX, attr);
961 proto = ip_conntrack_proto_find_get(npt);
963 if (proto->from_nfattr)
964 err = proto->from_nfattr(tb, ct);
965 ip_conntrack_proto_put(proto);
971 ctnetlink_change_conntrack(struct ip_conntrack *ct, struct nfattr *cda[])
975 DEBUGP("entered %s\n", __FUNCTION__);
977 if (cda[CTA_HELP-1]) {
978 err = ctnetlink_change_helper(ct, cda);
983 if (cda[CTA_TIMEOUT-1]) {
984 err = ctnetlink_change_timeout(ct, cda);
989 if (cda[CTA_STATUS-1]) {
990 err = ctnetlink_change_status(ct, cda);
995 if (cda[CTA_PROTOINFO-1]) {
996 err = ctnetlink_change_protoinfo(ct, cda);
1001 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1002 if (cda[CTA_MARK-1])
1003 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1006 DEBUGP("all done\n");
1011 ctnetlink_create_conntrack(struct nfattr *cda[],
1012 struct ip_conntrack_tuple *otuple,
1013 struct ip_conntrack_tuple *rtuple)
1015 struct ip_conntrack *ct;
1018 DEBUGP("entered %s\n", __FUNCTION__);
1020 ct = ip_conntrack_alloc(otuple, rtuple);
1021 if (ct == NULL || IS_ERR(ct))
1024 if (!cda[CTA_TIMEOUT-1])
1026 ct->timeout.expires = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_TIMEOUT-1]));
1028 ct->timeout.expires = jiffies + ct->timeout.expires * HZ;
1029 ct->status |= IPS_CONFIRMED;
1031 err = ctnetlink_change_status(ct, cda);
1035 if (cda[CTA_PROTOINFO-1]) {
1036 err = ctnetlink_change_protoinfo(ct, cda);
1041 ct->helper = ip_conntrack_helper_find_get(rtuple);
1043 add_timer(&ct->timeout);
1044 ip_conntrack_hash_insert(ct);
1047 ip_conntrack_helper_put(ct->helper);
1049 #if defined(CONFIG_IP_NF_CONNTRACK_MARK)
1050 if (cda[CTA_MARK-1])
1051 ct->mark = ntohl(*(u_int32_t *)NFA_DATA(cda[CTA_MARK-1]));
1054 DEBUGP("conntrack with id %u inserted\n", ct->id);
1058 ip_conntrack_free(ct);
1063 ctnetlink_new_conntrack(struct sock *ctnl, struct sk_buff *skb,
1064 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1066 struct ip_conntrack_tuple otuple, rtuple;
1067 struct ip_conntrack_tuple_hash *h = NULL;
1070 DEBUGP("entered %s\n", __FUNCTION__);
1072 if (nfattr_bad_size(cda, CTA_MAX, cta_min))
1075 if (cda[CTA_TUPLE_ORIG-1]) {
1076 err = ctnetlink_parse_tuple(cda, &otuple, CTA_TUPLE_ORIG);
1081 if (cda[CTA_TUPLE_REPLY-1]) {
1082 err = ctnetlink_parse_tuple(cda, &rtuple, CTA_TUPLE_REPLY);
1087 write_lock_bh(&ip_conntrack_lock);
1088 if (cda[CTA_TUPLE_ORIG-1])
1089 h = __ip_conntrack_find(&otuple, NULL);
1090 else if (cda[CTA_TUPLE_REPLY-1])
1091 h = __ip_conntrack_find(&rtuple, NULL);
1094 write_unlock_bh(&ip_conntrack_lock);
1095 DEBUGP("no such conntrack, create new\n");
1097 if (nlh->nlmsg_flags & NLM_F_CREATE)
1098 err = ctnetlink_create_conntrack(cda, &otuple, &rtuple);
1101 /* implicit 'else' */
1103 /* we only allow nat config for new conntracks */
1104 if (cda[CTA_NAT-1]) {
1109 /* We manipulate the conntrack inside the global conntrack table lock,
1110 * so there's no need to increase the refcount */
1111 DEBUGP("conntrack found\n");
1113 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1114 err = ctnetlink_change_conntrack(tuplehash_to_ctrack(h), cda);
1117 write_unlock_bh(&ip_conntrack_lock);
1121 /***********************************************************************
1123 ***********************************************************************/
1126 ctnetlink_exp_dump_tuple(struct sk_buff *skb,
1127 const struct ip_conntrack_tuple *tuple,
1128 enum ctattr_expect type)
1130 struct nfattr *nest_parms = NFA_NEST(skb, type);
1132 if (ctnetlink_dump_tuples(skb, tuple) < 0)
1133 goto nfattr_failure;
1135 NFA_NEST_END(skb, nest_parms);
1144 ctnetlink_exp_dump_expect(struct sk_buff *skb,
1145 const struct ip_conntrack_expect *exp)
1147 struct ip_conntrack *master = exp->master;
1148 u_int32_t timeout = htonl((exp->timeout.expires - jiffies) / HZ);
1149 u_int32_t id = htonl(exp->id);
1151 if (ctnetlink_exp_dump_tuple(skb, &exp->tuple, CTA_EXPECT_TUPLE) < 0)
1152 goto nfattr_failure;
1153 if (ctnetlink_exp_dump_tuple(skb, &exp->mask, CTA_EXPECT_MASK) < 0)
1154 goto nfattr_failure;
1155 if (ctnetlink_exp_dump_tuple(skb,
1156 &master->tuplehash[IP_CT_DIR_ORIGINAL].tuple,
1157 CTA_EXPECT_MASTER) < 0)
1158 goto nfattr_failure;
1160 NFA_PUT(skb, CTA_EXPECT_TIMEOUT, sizeof(timeout), &timeout);
1161 NFA_PUT(skb, CTA_EXPECT_ID, sizeof(u_int32_t), &id);
1170 ctnetlink_exp_fill_info(struct sk_buff *skb, u32 pid, u32 seq,
1173 const struct ip_conntrack_expect *exp)
1175 struct nlmsghdr *nlh;
1176 struct nfgenmsg *nfmsg;
1181 event |= NFNL_SUBSYS_CTNETLINK_EXP << 8;
1182 nlh = NLMSG_PUT(skb, pid, seq, event, sizeof(struct nfgenmsg));
1183 nfmsg = NLMSG_DATA(nlh);
1185 nlh->nlmsg_flags = (nowait && pid) ? NLM_F_MULTI : 0;
1186 nfmsg->nfgen_family = AF_INET;
1187 nfmsg->version = NFNETLINK_V0;
1190 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1191 goto nfattr_failure;
1193 nlh->nlmsg_len = skb->tail - b;
1198 skb_trim(skb, b - skb->data);
1202 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1203 static int ctnetlink_expect_event(struct notifier_block *this,
1204 unsigned long events, void *ptr)
1206 struct nlmsghdr *nlh;
1207 struct nfgenmsg *nfmsg;
1208 struct ip_conntrack_expect *exp = (struct ip_conntrack_expect *)ptr;
1209 struct sk_buff *skb;
1215 if (events & IPEXP_NEW) {
1216 type = IPCTNL_MSG_EXP_NEW;
1217 flags = NLM_F_CREATE|NLM_F_EXCL;
1221 skb = alloc_skb(NLMSG_GOODSIZE, GFP_ATOMIC);
1227 type |= NFNL_SUBSYS_CTNETLINK << 8;
1228 nlh = NLMSG_PUT(skb, 0, 0, type, sizeof(struct nfgenmsg));
1229 nfmsg = NLMSG_DATA(nlh);
1231 nlh->nlmsg_flags = flags;
1232 nfmsg->nfgen_family = AF_INET;
1233 nfmsg->version = NFNETLINK_V0;
1236 if (ctnetlink_exp_dump_expect(skb, exp) < 0)
1237 goto nfattr_failure;
1239 nlh->nlmsg_len = skb->tail - b;
1240 proto = exp->tuple.dst.protonum;
1241 nfnetlink_send(skb, 0, NFNLGRP_CONNTRACK_EXP_NEW, 0);
1252 ctnetlink_exp_dump_table(struct sk_buff *skb, struct netlink_callback *cb)
1254 struct ip_conntrack_expect *exp = NULL;
1255 struct list_head *i;
1256 u_int32_t *id = (u_int32_t *) &cb->args[0];
1258 DEBUGP("entered %s, last id=%llu\n", __FUNCTION__, *id);
1260 read_lock_bh(&ip_conntrack_lock);
1261 list_for_each_prev(i, &ip_conntrack_expect_list) {
1262 exp = (struct ip_conntrack_expect *) i;
1265 if (ctnetlink_exp_fill_info(skb, NETLINK_CB(cb->skb).pid,
1273 read_unlock_bh(&ip_conntrack_lock);
1275 DEBUGP("leaving, last id=%llu\n", *id);
1280 static const size_t cta_min_exp[CTA_EXPECT_MAX] = {
1281 [CTA_EXPECT_TIMEOUT-1] = sizeof(u_int32_t),
1282 [CTA_EXPECT_ID-1] = sizeof(u_int32_t)
1286 ctnetlink_get_expect(struct sock *ctnl, struct sk_buff *skb,
1287 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1289 struct ip_conntrack_tuple tuple;
1290 struct ip_conntrack_expect *exp;
1291 struct sk_buff *skb2;
1294 DEBUGP("entered %s\n", __FUNCTION__);
1296 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1299 if (nlh->nlmsg_flags & NLM_F_DUMP) {
1300 struct nfgenmsg *msg = NLMSG_DATA(nlh);
1303 if (msg->nfgen_family != AF_INET)
1304 return -EAFNOSUPPORT;
1306 if ((*errp = netlink_dump_start(ctnl, skb, nlh,
1307 ctnetlink_exp_dump_table,
1308 ctnetlink_done)) != 0)
1310 rlen = NLMSG_ALIGN(nlh->nlmsg_len);
1311 if (rlen > skb->len)
1313 skb_pull(skb, rlen);
1317 if (cda[CTA_EXPECT_MASTER-1])
1318 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_MASTER);
1325 exp = ip_conntrack_expect_find(&tuple);
1329 if (cda[CTA_EXPECT_ID-1]) {
1330 u_int32_t id = *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1331 if (exp->id != ntohl(id)) {
1332 ip_conntrack_expect_put(exp);
1338 skb2 = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
1341 NETLINK_CB(skb2).dst_pid = NETLINK_CB(skb).pid;
1343 err = ctnetlink_exp_fill_info(skb2, NETLINK_CB(skb).pid,
1344 nlh->nlmsg_seq, IPCTNL_MSG_EXP_NEW,
1349 ip_conntrack_expect_put(exp);
1351 return netlink_unicast(ctnl, skb2, NETLINK_CB(skb).pid, MSG_DONTWAIT);
1356 ip_conntrack_expect_put(exp);
1361 ctnetlink_del_expect(struct sock *ctnl, struct sk_buff *skb,
1362 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1364 struct ip_conntrack_expect *exp, *tmp;
1365 struct ip_conntrack_tuple tuple;
1366 struct ip_conntrack_helper *h;
1369 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1372 if (cda[CTA_EXPECT_TUPLE-1]) {
1373 /* delete a single expect by tuple */
1374 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1378 /* bump usage count to 2 */
1379 exp = ip_conntrack_expect_find(&tuple);
1383 if (cda[CTA_EXPECT_ID-1]) {
1385 *(u_int32_t *)NFA_DATA(cda[CTA_EXPECT_ID-1]);
1386 if (exp->id != ntohl(id)) {
1387 ip_conntrack_expect_put(exp);
1392 /* after list removal, usage count == 1 */
1393 ip_conntrack_unexpect_related(exp);
1394 /* have to put what we 'get' above.
1395 * after this line usage count == 0 */
1396 ip_conntrack_expect_put(exp);
1397 } else if (cda[CTA_EXPECT_HELP_NAME-1]) {
1398 char *name = NFA_DATA(cda[CTA_EXPECT_HELP_NAME-1]);
1400 /* delete all expectations for this helper */
1401 write_lock_bh(&ip_conntrack_lock);
1402 h = __ip_conntrack_helper_find_byname(name);
1404 write_unlock_bh(&ip_conntrack_lock);
1407 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1409 if (exp->master->helper == h
1410 && del_timer(&exp->timeout)) {
1411 ip_ct_unlink_expect(exp);
1412 ip_conntrack_expect_put(exp);
1415 write_unlock_bh(&ip_conntrack_lock);
1417 /* This basically means we have to flush everything*/
1418 write_lock_bh(&ip_conntrack_lock);
1419 list_for_each_entry_safe(exp, tmp, &ip_conntrack_expect_list,
1421 if (del_timer(&exp->timeout)) {
1422 ip_ct_unlink_expect(exp);
1423 ip_conntrack_expect_put(exp);
1426 write_unlock_bh(&ip_conntrack_lock);
1432 ctnetlink_change_expect(struct ip_conntrack_expect *x, struct nfattr *cda[])
1438 ctnetlink_create_expect(struct nfattr *cda[])
1440 struct ip_conntrack_tuple tuple, mask, master_tuple;
1441 struct ip_conntrack_tuple_hash *h = NULL;
1442 struct ip_conntrack_expect *exp;
1443 struct ip_conntrack *ct;
1446 DEBUGP("entered %s\n", __FUNCTION__);
1448 /* caller guarantees that those three CTA_EXPECT_* exist */
1449 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1452 err = ctnetlink_parse_tuple(cda, &mask, CTA_EXPECT_MASK);
1455 err = ctnetlink_parse_tuple(cda, &master_tuple, CTA_EXPECT_MASTER);
1459 /* Look for master conntrack of this expectation */
1460 h = ip_conntrack_find_get(&master_tuple, NULL);
1463 ct = tuplehash_to_ctrack(h);
1466 /* such conntrack hasn't got any helper, abort */
1471 exp = ip_conntrack_expect_alloc(ct);
1477 exp->expectfn = NULL;
1480 memcpy(&exp->tuple, &tuple, sizeof(struct ip_conntrack_tuple));
1481 memcpy(&exp->mask, &mask, sizeof(struct ip_conntrack_tuple));
1483 err = ip_conntrack_expect_related(exp);
1484 ip_conntrack_expect_put(exp);
1487 ip_conntrack_put(tuplehash_to_ctrack(h));
1492 ctnetlink_new_expect(struct sock *ctnl, struct sk_buff *skb,
1493 struct nlmsghdr *nlh, struct nfattr *cda[], int *errp)
1495 struct ip_conntrack_tuple tuple;
1496 struct ip_conntrack_expect *exp;
1499 DEBUGP("entered %s\n", __FUNCTION__);
1501 if (nfattr_bad_size(cda, CTA_EXPECT_MAX, cta_min_exp))
1504 if (!cda[CTA_EXPECT_TUPLE-1]
1505 || !cda[CTA_EXPECT_MASK-1]
1506 || !cda[CTA_EXPECT_MASTER-1])
1509 err = ctnetlink_parse_tuple(cda, &tuple, CTA_EXPECT_TUPLE);
1513 write_lock_bh(&ip_conntrack_lock);
1514 exp = __ip_conntrack_expect_find(&tuple);
1517 write_unlock_bh(&ip_conntrack_lock);
1519 if (nlh->nlmsg_flags & NLM_F_CREATE)
1520 err = ctnetlink_create_expect(cda);
1525 if (!(nlh->nlmsg_flags & NLM_F_EXCL))
1526 err = ctnetlink_change_expect(exp, cda);
1527 write_unlock_bh(&ip_conntrack_lock);
1529 DEBUGP("leaving\n");
1534 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1535 static struct notifier_block ctnl_notifier = {
1536 .notifier_call = ctnetlink_conntrack_event,
1539 static struct notifier_block ctnl_notifier_exp = {
1540 .notifier_call = ctnetlink_expect_event,
1544 static struct nfnl_callback ctnl_cb[IPCTNL_MSG_MAX] = {
1545 [IPCTNL_MSG_CT_NEW] = { .call = ctnetlink_new_conntrack,
1546 .attr_count = CTA_MAX, },
1547 [IPCTNL_MSG_CT_GET] = { .call = ctnetlink_get_conntrack,
1548 .attr_count = CTA_MAX, },
1549 [IPCTNL_MSG_CT_DELETE] = { .call = ctnetlink_del_conntrack,
1550 .attr_count = CTA_MAX, },
1551 [IPCTNL_MSG_CT_GET_CTRZERO] = { .call = ctnetlink_get_conntrack,
1552 .attr_count = CTA_MAX, },
1555 static struct nfnl_callback ctnl_exp_cb[IPCTNL_MSG_EXP_MAX] = {
1556 [IPCTNL_MSG_EXP_GET] = { .call = ctnetlink_get_expect,
1557 .attr_count = CTA_EXPECT_MAX, },
1558 [IPCTNL_MSG_EXP_NEW] = { .call = ctnetlink_new_expect,
1559 .attr_count = CTA_EXPECT_MAX, },
1560 [IPCTNL_MSG_EXP_DELETE] = { .call = ctnetlink_del_expect,
1561 .attr_count = CTA_EXPECT_MAX, },
1564 static struct nfnetlink_subsystem ctnl_subsys = {
1565 .name = "conntrack",
1566 .subsys_id = NFNL_SUBSYS_CTNETLINK,
1567 .cb_count = IPCTNL_MSG_MAX,
1571 static struct nfnetlink_subsystem ctnl_exp_subsys = {
1572 .name = "conntrack_expect",
1573 .subsys_id = NFNL_SUBSYS_CTNETLINK_EXP,
1574 .cb_count = IPCTNL_MSG_EXP_MAX,
1578 MODULE_ALIAS_NFNL_SUBSYS(NFNL_SUBSYS_CTNETLINK);
1580 static int __init ctnetlink_init(void)
1584 printk("ctnetlink v%s: registering with nfnetlink.\n", version);
1585 ret = nfnetlink_subsys_register(&ctnl_subsys);
1587 printk("ctnetlink_init: cannot register with nfnetlink.\n");
1591 ret = nfnetlink_subsys_register(&ctnl_exp_subsys);
1593 printk("ctnetlink_init: cannot register exp with nfnetlink.\n");
1594 goto err_unreg_subsys;
1597 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1598 ret = ip_conntrack_register_notifier(&ctnl_notifier);
1600 printk("ctnetlink_init: cannot register notifier.\n");
1601 goto err_unreg_exp_subsys;
1604 ret = ip_conntrack_expect_register_notifier(&ctnl_notifier_exp);
1606 printk("ctnetlink_init: cannot expect register notifier.\n");
1607 goto err_unreg_notifier;
1613 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1615 ip_conntrack_unregister_notifier(&ctnl_notifier);
1616 err_unreg_exp_subsys:
1617 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1620 nfnetlink_subsys_unregister(&ctnl_subsys);
1625 static void __exit ctnetlink_exit(void)
1627 printk("ctnetlink: unregistering from nfnetlink.\n");
1629 #ifdef CONFIG_IP_NF_CONNTRACK_EVENTS
1630 ip_conntrack_unregister_notifier(&ctnl_notifier_exp);
1631 ip_conntrack_unregister_notifier(&ctnl_notifier);
1634 nfnetlink_subsys_unregister(&ctnl_exp_subsys);
1635 nfnetlink_subsys_unregister(&ctnl_subsys);
1639 module_init(ctnetlink_init);
1640 module_exit(ctnetlink_exit);