1 // SPDX-License-Identifier: GPL-2.0 OR Linux-OpenIB
3 * net/sched/act_ct.c Connection Tracking action
5 * Authors: Paul Blakey <paulb@mellanox.com>
6 * Yossi Kuperman <yossiku@mellanox.com>
7 * Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
10 #include <linux/module.h>
11 #include <linux/init.h>
12 #include <linux/kernel.h>
13 #include <linux/skbuff.h>
14 #include <linux/rtnetlink.h>
15 #include <linux/pkt_cls.h>
17 #include <linux/ipv6.h>
18 #include <net/netlink.h>
19 #include <net/pkt_sched.h>
20 #include <net/pkt_cls.h>
21 #include <net/act_api.h>
23 #include <net/ipv6_frag.h>
24 #include <uapi/linux/tc_act/tc_ct.h>
25 #include <net/tc_act/tc_ct.h>
27 #include <net/netfilter/nf_conntrack.h>
28 #include <net/netfilter/nf_conntrack_core.h>
29 #include <net/netfilter/nf_conntrack_zones.h>
30 #include <net/netfilter/nf_conntrack_helper.h>
31 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
32 #include <uapi/linux/netfilter/nf_nat.h>
34 static struct tc_action_ops act_ct_ops;
35 static unsigned int ct_net_id;
37 struct tc_ct_action_net {
38 struct tc_action_net tn; /* Must be first */
42 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
43 static bool tcf_ct_skb_nfct_cached(struct net *net, struct sk_buff *skb,
44 u16 zone_id, bool force)
46 enum ip_conntrack_info ctinfo;
49 ct = nf_ct_get(skb, &ctinfo);
52 if (!net_eq(net, read_pnet(&ct->ct_net)))
54 if (nf_ct_zone(ct)->id != zone_id)
57 /* Force conntrack entry direction. */
58 if (force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
59 if (nf_ct_is_confirmed(ct))
62 nf_conntrack_put(&ct->ct_general);
63 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
71 /* Trim the skb to the length specified by the IP/IPv6 header,
72 * removing any trailing lower-layer padding. This prepares the skb
73 * for higher-layer processing that assumes skb->len excludes padding
74 * (such as nf_ip_checksum). The caller needs to pull the skb to the
75 * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
77 static int tcf_ct_skb_network_trim(struct sk_buff *skb, int family)
84 len = ntohs(ip_hdr(skb)->tot_len);
87 len = sizeof(struct ipv6hdr)
88 + ntohs(ipv6_hdr(skb)->payload_len);
94 err = pskb_trim_rcsum(skb, len);
99 static u8 tcf_ct_skb_nf_family(struct sk_buff *skb)
101 u8 family = NFPROTO_UNSPEC;
103 switch (skb_protocol(skb, true)) {
104 case htons(ETH_P_IP):
105 family = NFPROTO_IPV4;
107 case htons(ETH_P_IPV6):
108 family = NFPROTO_IPV6;
117 static int tcf_ct_ipv4_is_fragment(struct sk_buff *skb, bool *frag)
121 len = skb_network_offset(skb) + sizeof(struct iphdr);
122 if (unlikely(skb->len < len))
124 if (unlikely(!pskb_may_pull(skb, len)))
127 *frag = ip_is_fragment(ip_hdr(skb));
131 static int tcf_ct_ipv6_is_fragment(struct sk_buff *skb, bool *frag)
133 unsigned int flags = 0, len, payload_ofs = 0;
134 unsigned short frag_off;
137 len = skb_network_offset(skb) + sizeof(struct ipv6hdr);
138 if (unlikely(skb->len < len))
140 if (unlikely(!pskb_may_pull(skb, len)))
143 nexthdr = ipv6_find_hdr(skb, &payload_ofs, -1, &frag_off, &flags);
144 if (unlikely(nexthdr < 0))
147 *frag = flags & IP6_FH_F_FRAG;
151 static int tcf_ct_handle_fragments(struct net *net, struct sk_buff *skb,
154 enum ip_conntrack_info ctinfo;
159 /* Previously seen (loopback)? Ignore. */
160 ct = nf_ct_get(skb, &ctinfo);
161 if ((ct && !nf_ct_is_template(ct)) || ctinfo == IP_CT_UNTRACKED)
164 if (family == NFPROTO_IPV4)
165 err = tcf_ct_ipv4_is_fragment(skb, &frag);
167 err = tcf_ct_ipv6_is_fragment(skb, &frag);
173 if (family == NFPROTO_IPV4) {
174 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
176 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
178 err = ip_defrag(net, skb, user);
180 if (err && err != -EINPROGRESS)
182 } else { /* NFPROTO_IPV6 */
183 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
184 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
186 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
187 err = nf_ct_frag6_gather(net, skb, user);
188 if (err && err != -EINPROGRESS)
205 static void tcf_ct_params_free(struct rcu_head *head)
207 struct tcf_ct_params *params = container_of(head,
208 struct tcf_ct_params, rcu);
211 nf_conntrack_put(¶ms->tmpl->ct_general);
215 #if IS_ENABLED(CONFIG_NF_NAT)
216 /* Modelled after nf_nat_ipv[46]_fn().
217 * range is only used for new, uninitialized NAT state.
218 * Returns either NF_ACCEPT or NF_DROP.
220 static int ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
221 enum ip_conntrack_info ctinfo,
222 const struct nf_nat_range2 *range,
223 enum nf_nat_manip_type maniptype)
225 __be16 proto = skb_protocol(skb, true);
226 int hooknum, err = NF_ACCEPT;
228 /* See HOOK2MANIP(). */
229 if (maniptype == NF_NAT_MANIP_SRC)
230 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
232 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
236 case IP_CT_RELATED_REPLY:
237 if (proto == htons(ETH_P_IP) &&
238 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
239 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
243 } else if (IS_ENABLED(CONFIG_IPV6) && proto == htons(ETH_P_IPV6)) {
245 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
246 int hdrlen = ipv6_skip_exthdr(skb,
247 sizeof(struct ipv6hdr),
248 &nexthdr, &frag_off);
250 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
251 if (!nf_nat_icmpv6_reply_translation(skb, ct,
259 /* Non-ICMP, fall thru to initialize if needed. */
262 /* Seen it before? This can happen for loopback, retrans,
265 if (!nf_nat_initialized(ct, maniptype)) {
266 /* Initialize according to the NAT action. */
267 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
268 /* Action is set up to establish a new
271 ? nf_nat_setup_info(ct, range, maniptype)
272 : nf_nat_alloc_null_binding(ct, hooknum);
273 if (err != NF_ACCEPT)
278 case IP_CT_ESTABLISHED:
279 case IP_CT_ESTABLISHED_REPLY:
287 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
291 #endif /* CONFIG_NF_NAT */
293 static void tcf_ct_act_set_mark(struct nf_conn *ct, u32 mark, u32 mask)
295 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
301 new_mark = mark | (ct->mark & ~(mask));
302 if (ct->mark != new_mark) {
304 if (nf_ct_is_confirmed(ct))
305 nf_conntrack_event_cache(IPCT_MARK, ct);
310 static void tcf_ct_act_set_labels(struct nf_conn *ct,
314 #if IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)
315 size_t labels_sz = FIELD_SIZEOF(struct tcf_ct_params, labels);
317 if (!memchr_inv(labels_m, 0, labels_sz))
320 nf_connlabels_replace(ct, labels, labels_m, 4);
324 static int tcf_ct_act_nat(struct sk_buff *skb,
326 enum ip_conntrack_info ctinfo,
328 struct nf_nat_range2 *range,
331 #if IS_ENABLED(CONFIG_NF_NAT)
333 enum nf_nat_manip_type maniptype;
335 if (!(ct_action & TCA_CT_ACT_NAT))
338 /* Add NAT extension if not confirmed yet. */
339 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
340 return NF_DROP; /* Can't NAT. */
342 if (ctinfo != IP_CT_NEW && (ct->status & IPS_NAT_MASK) &&
343 (ctinfo != IP_CT_RELATED || commit)) {
344 /* NAT an established or related connection like before. */
345 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
346 /* This is the REPLY direction for a connection
347 * for which NAT was applied in the forward
348 * direction. Do the reverse NAT.
350 maniptype = ct->status & IPS_SRC_NAT
351 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
353 maniptype = ct->status & IPS_SRC_NAT
354 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
355 } else if (ct_action & TCA_CT_ACT_NAT_SRC) {
356 maniptype = NF_NAT_MANIP_SRC;
357 } else if (ct_action & TCA_CT_ACT_NAT_DST) {
358 maniptype = NF_NAT_MANIP_DST;
363 err = ct_nat_execute(skb, ct, ctinfo, range, maniptype);
364 if (err == NF_ACCEPT && ct->status & IPS_DST_NAT) {
365 if (ct->status & IPS_SRC_NAT) {
366 if (maniptype == NF_NAT_MANIP_SRC)
367 maniptype = NF_NAT_MANIP_DST;
369 maniptype = NF_NAT_MANIP_SRC;
371 err = ct_nat_execute(skb, ct, ctinfo, range,
373 } else if (CTINFO2DIR(ctinfo) == IP_CT_DIR_ORIGINAL) {
374 err = ct_nat_execute(skb, ct, ctinfo, NULL,
384 static int tcf_ct_act(struct sk_buff *skb, const struct tc_action *a,
385 struct tcf_result *res)
387 struct net *net = dev_net(skb->dev);
388 bool cached, commit, clear, force;
389 enum ip_conntrack_info ctinfo;
390 struct tcf_ct *c = to_ct(a);
391 struct nf_conn *tmpl = NULL;
392 struct nf_hook_state state;
393 int nh_ofs, err, retval;
394 struct tcf_ct_params *p;
398 p = rcu_dereference_bh(c->params);
400 retval = READ_ONCE(c->tcf_action);
401 commit = p->ct_action & TCA_CT_ACT_COMMIT;
402 clear = p->ct_action & TCA_CT_ACT_CLEAR;
403 force = p->ct_action & TCA_CT_ACT_FORCE;
407 ct = nf_ct_get(skb, &ctinfo);
409 nf_conntrack_put(&ct->ct_general);
410 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
416 family = tcf_ct_skb_nf_family(skb);
417 if (family == NFPROTO_UNSPEC)
420 /* The conntrack module expects to be working at L3.
421 * We also try to pull the IPv4/6 header to linear area
423 nh_ofs = skb_network_offset(skb);
424 skb_pull_rcsum(skb, nh_ofs);
425 err = tcf_ct_handle_fragments(net, skb, family, p->zone);
426 if (err == -EINPROGRESS) {
427 retval = TC_ACT_STOLEN;
433 err = tcf_ct_skb_network_trim(skb, family);
437 /* If we are recirculating packets to match on ct fields and
438 * committing with a separate ct action, then we don't need to
439 * actually run the packet through conntrack twice unless it's for a
442 cached = tcf_ct_skb_nfct_cached(net, skb, p->zone, force);
444 /* Associate skb with specified zone. */
446 ct = nf_ct_get(skb, &ctinfo);
448 nf_conntrack_put(skb_nfct(skb));
449 nf_conntrack_get(&tmpl->ct_general);
450 nf_ct_set(skb, tmpl, IP_CT_NEW);
453 state.hook = NF_INET_PRE_ROUTING;
456 err = nf_conntrack_in(skb, &state);
457 if (err != NF_ACCEPT)
461 ct = nf_ct_get(skb, &ctinfo);
464 nf_ct_deliver_cached_events(ct);
466 err = tcf_ct_act_nat(skb, ct, ctinfo, p->ct_action, &p->range, commit);
467 if (err != NF_ACCEPT)
471 tcf_ct_act_set_mark(ct, p->mark, p->mark_mask);
472 tcf_ct_act_set_labels(ct, p->labels, p->labels_mask);
474 /* This will take care of sending queued events
475 * even if the connection is already confirmed.
477 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
482 skb_push_rcsum(skb, nh_ofs);
485 bstats_cpu_update(this_cpu_ptr(a->cpu_bstats), skb);
489 qstats_drop_inc(this_cpu_ptr(a->cpu_qstats));
493 static const struct nla_policy ct_policy[TCA_CT_MAX + 1] = {
494 [TCA_CT_UNSPEC] = { .strict_start_type = TCA_CT_UNSPEC + 1 },
495 [TCA_CT_ACTION] = { .type = NLA_U16 },
496 [TCA_CT_PARMS] = { .type = NLA_EXACT_LEN, .len = sizeof(struct tc_ct) },
497 [TCA_CT_ZONE] = { .type = NLA_U16 },
498 [TCA_CT_MARK] = { .type = NLA_U32 },
499 [TCA_CT_MARK_MASK] = { .type = NLA_U32 },
500 [TCA_CT_LABELS] = { .type = NLA_BINARY,
501 .len = 128 / BITS_PER_BYTE },
502 [TCA_CT_LABELS_MASK] = { .type = NLA_BINARY,
503 .len = 128 / BITS_PER_BYTE },
504 [TCA_CT_NAT_IPV4_MIN] = { .type = NLA_U32 },
505 [TCA_CT_NAT_IPV4_MAX] = { .type = NLA_U32 },
506 [TCA_CT_NAT_IPV6_MIN] = { .type = NLA_EXACT_LEN,
507 .len = sizeof(struct in6_addr) },
508 [TCA_CT_NAT_IPV6_MAX] = { .type = NLA_EXACT_LEN,
509 .len = sizeof(struct in6_addr) },
510 [TCA_CT_NAT_PORT_MIN] = { .type = NLA_U16 },
511 [TCA_CT_NAT_PORT_MAX] = { .type = NLA_U16 },
514 static int tcf_ct_fill_params_nat(struct tcf_ct_params *p,
517 struct netlink_ext_ack *extack)
519 struct nf_nat_range2 *range;
521 if (!(p->ct_action & TCA_CT_ACT_NAT))
524 if (!IS_ENABLED(CONFIG_NF_NAT)) {
525 NL_SET_ERR_MSG_MOD(extack, "Netfilter nat isn't enabled in kernel");
529 if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
532 if ((p->ct_action & TCA_CT_ACT_NAT_SRC) &&
533 (p->ct_action & TCA_CT_ACT_NAT_DST)) {
534 NL_SET_ERR_MSG_MOD(extack, "dnat and snat can't be enabled at the same time");
539 if (tb[TCA_CT_NAT_IPV4_MIN]) {
540 struct nlattr *max_attr = tb[TCA_CT_NAT_IPV4_MAX];
542 p->ipv4_range = true;
543 range->flags |= NF_NAT_RANGE_MAP_IPS;
545 nla_get_in_addr(tb[TCA_CT_NAT_IPV4_MIN]);
547 range->max_addr.ip = max_attr ?
548 nla_get_in_addr(max_attr) :
550 } else if (tb[TCA_CT_NAT_IPV6_MIN]) {
551 struct nlattr *max_attr = tb[TCA_CT_NAT_IPV6_MAX];
553 p->ipv4_range = false;
554 range->flags |= NF_NAT_RANGE_MAP_IPS;
555 range->min_addr.in6 =
556 nla_get_in6_addr(tb[TCA_CT_NAT_IPV6_MIN]);
558 range->max_addr.in6 = max_attr ?
559 nla_get_in6_addr(max_attr) :
563 if (tb[TCA_CT_NAT_PORT_MIN]) {
564 range->flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
565 range->min_proto.all = nla_get_be16(tb[TCA_CT_NAT_PORT_MIN]);
567 range->max_proto.all = tb[TCA_CT_NAT_PORT_MAX] ?
568 nla_get_be16(tb[TCA_CT_NAT_PORT_MAX]) :
569 range->min_proto.all;
575 static void tcf_ct_set_key_val(struct nlattr **tb,
576 void *val, int val_type,
577 void *mask, int mask_type,
582 nla_memcpy(val, tb[val_type], len);
587 if (mask_type == TCA_CT_UNSPEC || !tb[mask_type])
588 memset(mask, 0xff, len);
590 nla_memcpy(mask, tb[mask_type], len);
593 static int tcf_ct_fill_params(struct net *net,
594 struct tcf_ct_params *p,
597 struct netlink_ext_ack *extack)
599 struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
600 struct nf_conntrack_zone zone;
601 struct nf_conn *tmpl;
604 p->zone = NF_CT_DEFAULT_ZONE_ID;
606 tcf_ct_set_key_val(tb,
607 &p->ct_action, TCA_CT_ACTION,
609 sizeof(p->ct_action));
611 if (p->ct_action & TCA_CT_ACT_CLEAR)
614 err = tcf_ct_fill_params_nat(p, parm, tb, extack);
618 if (tb[TCA_CT_MARK]) {
619 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
620 NL_SET_ERR_MSG_MOD(extack, "Conntrack mark isn't enabled.");
623 tcf_ct_set_key_val(tb,
624 &p->mark, TCA_CT_MARK,
625 &p->mark_mask, TCA_CT_MARK_MASK,
629 if (tb[TCA_CT_LABELS]) {
630 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
631 NL_SET_ERR_MSG_MOD(extack, "Conntrack labels isn't enabled.");
636 NL_SET_ERR_MSG_MOD(extack, "Failed to set connlabel length");
639 tcf_ct_set_key_val(tb,
640 p->labels, TCA_CT_LABELS,
641 p->labels_mask, TCA_CT_LABELS_MASK,
645 if (tb[TCA_CT_ZONE]) {
646 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
647 NL_SET_ERR_MSG_MOD(extack, "Conntrack zones isn't enabled.");
651 tcf_ct_set_key_val(tb,
652 &p->zone, TCA_CT_ZONE,
657 nf_ct_zone_init(&zone, p->zone, NF_CT_DEFAULT_ZONE_DIR, 0);
658 tmpl = nf_ct_tmpl_alloc(net, &zone, GFP_KERNEL);
660 NL_SET_ERR_MSG_MOD(extack, "Failed to allocate conntrack template");
663 __set_bit(IPS_CONFIRMED_BIT, &tmpl->status);
664 nf_conntrack_get(&tmpl->ct_general);
670 static int tcf_ct_init(struct net *net, struct nlattr *nla,
671 struct nlattr *est, struct tc_action **a,
672 int replace, int bind, bool rtnl_held,
673 struct tcf_proto *tp,
674 struct netlink_ext_ack *extack)
676 struct tc_action_net *tn = net_generic(net, ct_net_id);
677 struct tcf_ct_params *params = NULL;
678 struct nlattr *tb[TCA_CT_MAX + 1];
679 struct tcf_chain *goto_ch = NULL;
686 NL_SET_ERR_MSG_MOD(extack, "Ct requires attributes to be passed");
690 err = nla_parse_nested(tb, TCA_CT_MAX, nla, ct_policy, extack);
694 if (!tb[TCA_CT_PARMS]) {
695 NL_SET_ERR_MSG_MOD(extack, "Missing required ct parameters");
698 parm = nla_data(tb[TCA_CT_PARMS]);
700 err = tcf_idr_check_alloc(tn, &index, a, bind);
705 err = tcf_idr_create(tn, index, est, a,
706 &act_ct_ops, bind, true);
708 tcf_idr_cleanup(tn, index);
717 tcf_idr_release(*a, bind);
721 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
727 params = kzalloc(sizeof(*params), GFP_KERNEL);
728 if (unlikely(!params)) {
733 err = tcf_ct_fill_params(net, params, parm, tb, extack);
737 spin_lock_bh(&c->tcf_lock);
738 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
739 rcu_swap_protected(c->params, params, lockdep_is_held(&c->tcf_lock));
740 spin_unlock_bh(&c->tcf_lock);
743 tcf_chain_put_by_act(goto_ch);
745 call_rcu(¶ms->rcu, tcf_ct_params_free);
751 tcf_chain_put_by_act(goto_ch);
753 tcf_idr_release(*a, bind);
757 static void tcf_ct_cleanup(struct tc_action *a)
759 struct tcf_ct_params *params;
760 struct tcf_ct *c = to_ct(a);
762 params = rcu_dereference_protected(c->params, 1);
764 call_rcu(¶ms->rcu, tcf_ct_params_free);
767 static int tcf_ct_dump_key_val(struct sk_buff *skb,
768 void *val, int val_type,
769 void *mask, int mask_type,
774 if (mask && !memchr_inv(mask, 0, len))
777 err = nla_put(skb, val_type, len, val);
781 if (mask_type != TCA_CT_UNSPEC) {
782 err = nla_put(skb, mask_type, len, mask);
790 static int tcf_ct_dump_nat(struct sk_buff *skb, struct tcf_ct_params *p)
792 struct nf_nat_range2 *range = &p->range;
794 if (!(p->ct_action & TCA_CT_ACT_NAT))
797 if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
800 if (range->flags & NF_NAT_RANGE_MAP_IPS) {
802 if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MIN,
805 if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MAX,
809 if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MIN,
810 &range->min_addr.in6))
812 if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MAX,
813 &range->max_addr.in6))
818 if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) {
819 if (nla_put_be16(skb, TCA_CT_NAT_PORT_MIN,
820 range->min_proto.all))
822 if (nla_put_be16(skb, TCA_CT_NAT_PORT_MAX,
823 range->max_proto.all))
830 static inline int tcf_ct_dump(struct sk_buff *skb, struct tc_action *a,
833 unsigned char *b = skb_tail_pointer(skb);
834 struct tcf_ct *c = to_ct(a);
835 struct tcf_ct_params *p;
838 .index = c->tcf_index,
839 .refcnt = refcount_read(&c->tcf_refcnt) - ref,
840 .bindcnt = atomic_read(&c->tcf_bindcnt) - bind,
844 spin_lock_bh(&c->tcf_lock);
845 p = rcu_dereference_protected(c->params,
846 lockdep_is_held(&c->tcf_lock));
847 opt.action = c->tcf_action;
849 if (tcf_ct_dump_key_val(skb,
850 &p->ct_action, TCA_CT_ACTION,
852 sizeof(p->ct_action)))
853 goto nla_put_failure;
855 if (p->ct_action & TCA_CT_ACT_CLEAR)
858 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
859 tcf_ct_dump_key_val(skb,
860 &p->mark, TCA_CT_MARK,
861 &p->mark_mask, TCA_CT_MARK_MASK,
863 goto nla_put_failure;
865 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
866 tcf_ct_dump_key_val(skb,
867 p->labels, TCA_CT_LABELS,
868 p->labels_mask, TCA_CT_LABELS_MASK,
870 goto nla_put_failure;
872 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
873 tcf_ct_dump_key_val(skb,
874 &p->zone, TCA_CT_ZONE,
877 goto nla_put_failure;
879 if (tcf_ct_dump_nat(skb, p))
880 goto nla_put_failure;
883 if (nla_put(skb, TCA_CT_PARMS, sizeof(opt), &opt))
884 goto nla_put_failure;
886 tcf_tm_dump(&t, &c->tcf_tm);
887 if (nla_put_64bit(skb, TCA_CT_TM, sizeof(t), &t, TCA_CT_PAD))
888 goto nla_put_failure;
889 spin_unlock_bh(&c->tcf_lock);
893 spin_unlock_bh(&c->tcf_lock);
898 static int tcf_ct_walker(struct net *net, struct sk_buff *skb,
899 struct netlink_callback *cb, int type,
900 const struct tc_action_ops *ops,
901 struct netlink_ext_ack *extack)
903 struct tc_action_net *tn = net_generic(net, ct_net_id);
905 return tcf_generic_walker(tn, skb, cb, type, ops, extack);
908 static int tcf_ct_search(struct net *net, struct tc_action **a, u32 index)
910 struct tc_action_net *tn = net_generic(net, ct_net_id);
912 return tcf_idr_search(tn, a, index);
915 static void tcf_stats_update(struct tc_action *a, u64 bytes, u32 packets,
916 u64 lastuse, bool hw)
918 struct tcf_ct *c = to_ct(a);
920 _bstats_cpu_update(this_cpu_ptr(a->cpu_bstats), bytes, packets);
923 _bstats_cpu_update(this_cpu_ptr(a->cpu_bstats_hw),
925 c->tcf_tm.lastuse = max_t(u64, c->tcf_tm.lastuse, lastuse);
928 static struct tc_action_ops act_ct_ops = {
931 .owner = THIS_MODULE,
935 .cleanup = tcf_ct_cleanup,
936 .walk = tcf_ct_walker,
937 .lookup = tcf_ct_search,
938 .stats_update = tcf_stats_update,
939 .size = sizeof(struct tcf_ct),
942 static __net_init int ct_init_net(struct net *net)
944 unsigned int n_bits = FIELD_SIZEOF(struct tcf_ct_params, labels) * 8;
945 struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
947 if (nf_connlabels_get(net, n_bits - 1)) {
949 pr_err("act_ct: Failed to set connlabels length");
954 return tc_action_net_init(net, &tn->tn, &act_ct_ops);
957 static void __net_exit ct_exit_net(struct list_head *net_list)
962 list_for_each_entry(net, net_list, exit_list) {
963 struct tc_ct_action_net *tn = net_generic(net, ct_net_id);
966 nf_connlabels_put(net);
970 tc_action_net_exit(net_list, ct_net_id);
973 static struct pernet_operations ct_net_ops = {
975 .exit_batch = ct_exit_net,
977 .size = sizeof(struct tc_ct_action_net),
980 static int __init ct_init_module(void)
982 return tcf_register_action(&act_ct_ops, &ct_net_ops);
985 static void __exit ct_cleanup_module(void)
987 tcf_unregister_action(&act_ct_ops, &ct_net_ops);
990 module_init(ct_init_module);
991 module_exit(ct_cleanup_module);
992 MODULE_AUTHOR("Paul Blakey <paulb@mellanox.com>");
993 MODULE_AUTHOR("Yossi Kuperman <yossiku@mellanox.com>");
994 MODULE_AUTHOR("Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>");
995 MODULE_DESCRIPTION("Connection tracking action");
996 MODULE_LICENSE("GPL v2");