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 <linux/rhashtable.h>
19 #include <net/netlink.h>
20 #include <net/pkt_sched.h>
21 #include <net/pkt_cls.h>
22 #include <net/act_api.h>
24 #include <net/ipv6_frag.h>
25 #include <uapi/linux/tc_act/tc_ct.h>
26 #include <net/tc_act/tc_ct.h>
27 #include <net/tc_wrapper.h>
29 #include <net/netfilter/nf_flow_table.h>
30 #include <net/netfilter/nf_conntrack.h>
31 #include <net/netfilter/nf_conntrack_core.h>
32 #include <net/netfilter/nf_conntrack_zones.h>
33 #include <net/netfilter/nf_conntrack_helper.h>
34 #include <net/netfilter/nf_conntrack_acct.h>
35 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
36 #include <net/netfilter/nf_conntrack_act_ct.h>
37 #include <net/netfilter/nf_conntrack_seqadj.h>
38 #include <uapi/linux/netfilter/nf_nat.h>
40 static struct workqueue_struct *act_ct_wq;
41 static struct rhashtable zones_ht;
42 static DEFINE_MUTEX(zones_mutex);
44 struct tcf_ct_flow_table {
45 struct rhash_head node; /* In zones tables */
47 struct rcu_work rwork;
48 struct nf_flowtable nf_ft;
55 static const struct rhashtable_params zones_params = {
56 .head_offset = offsetof(struct tcf_ct_flow_table, node),
57 .key_offset = offsetof(struct tcf_ct_flow_table, zone),
58 .key_len = sizeof_field(struct tcf_ct_flow_table, zone),
59 .automatic_shrinking = true,
62 static struct flow_action_entry *
63 tcf_ct_flow_table_flow_action_get_next(struct flow_action *flow_action)
65 int i = flow_action->num_entries++;
67 return &flow_action->entries[i];
70 static void tcf_ct_add_mangle_action(struct flow_action *action,
71 enum flow_action_mangle_base htype,
76 struct flow_action_entry *entry;
78 entry = tcf_ct_flow_table_flow_action_get_next(action);
79 entry->id = FLOW_ACTION_MANGLE;
80 entry->mangle.htype = htype;
81 entry->mangle.mask = ~mask;
82 entry->mangle.offset = offset;
83 entry->mangle.val = val;
86 /* The following nat helper functions check if the inverted reverse tuple
87 * (target) is different then the current dir tuple - meaning nat for ports
88 * and/or ip is needed, and add the relevant mangle actions.
91 tcf_ct_flow_table_add_action_nat_ipv4(const struct nf_conntrack_tuple *tuple,
92 struct nf_conntrack_tuple target,
93 struct flow_action *action)
95 if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3)))
96 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4,
97 offsetof(struct iphdr, saddr),
99 be32_to_cpu(target.src.u3.ip));
100 if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3)))
101 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP4,
102 offsetof(struct iphdr, daddr),
104 be32_to_cpu(target.dst.u3.ip));
108 tcf_ct_add_ipv6_addr_mangle_action(struct flow_action *action,
109 union nf_inet_addr *addr,
114 for (i = 0; i < sizeof(struct in6_addr) / sizeof(u32); i++)
115 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_IP6,
116 i * sizeof(u32) + offset,
117 0xFFFFFFFF, be32_to_cpu(addr->ip6[i]));
121 tcf_ct_flow_table_add_action_nat_ipv6(const struct nf_conntrack_tuple *tuple,
122 struct nf_conntrack_tuple target,
123 struct flow_action *action)
125 if (memcmp(&target.src.u3, &tuple->src.u3, sizeof(target.src.u3)))
126 tcf_ct_add_ipv6_addr_mangle_action(action, &target.src.u3,
127 offsetof(struct ipv6hdr,
129 if (memcmp(&target.dst.u3, &tuple->dst.u3, sizeof(target.dst.u3)))
130 tcf_ct_add_ipv6_addr_mangle_action(action, &target.dst.u3,
131 offsetof(struct ipv6hdr,
136 tcf_ct_flow_table_add_action_nat_tcp(const struct nf_conntrack_tuple *tuple,
137 struct nf_conntrack_tuple target,
138 struct flow_action *action)
140 __be16 target_src = target.src.u.tcp.port;
141 __be16 target_dst = target.dst.u.tcp.port;
143 if (target_src != tuple->src.u.tcp.port)
144 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP,
145 offsetof(struct tcphdr, source),
146 0xFFFF, be16_to_cpu(target_src));
147 if (target_dst != tuple->dst.u.tcp.port)
148 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_TCP,
149 offsetof(struct tcphdr, dest),
150 0xFFFF, be16_to_cpu(target_dst));
154 tcf_ct_flow_table_add_action_nat_udp(const struct nf_conntrack_tuple *tuple,
155 struct nf_conntrack_tuple target,
156 struct flow_action *action)
158 __be16 target_src = target.src.u.udp.port;
159 __be16 target_dst = target.dst.u.udp.port;
161 if (target_src != tuple->src.u.udp.port)
162 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_UDP,
163 offsetof(struct udphdr, source),
164 0xFFFF, be16_to_cpu(target_src));
165 if (target_dst != tuple->dst.u.udp.port)
166 tcf_ct_add_mangle_action(action, FLOW_ACT_MANGLE_HDR_TYPE_UDP,
167 offsetof(struct udphdr, dest),
168 0xFFFF, be16_to_cpu(target_dst));
171 static void tcf_ct_flow_table_add_action_meta(struct nf_conn *ct,
172 enum ip_conntrack_dir dir,
173 enum ip_conntrack_info ctinfo,
174 struct flow_action *action)
176 struct nf_conn_labels *ct_labels;
177 struct flow_action_entry *entry;
180 entry = tcf_ct_flow_table_flow_action_get_next(action);
181 entry->id = FLOW_ACTION_CT_METADATA;
182 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
183 entry->ct_metadata.mark = READ_ONCE(ct->mark);
185 /* aligns with the CT reference on the SKB nf_ct_set */
186 entry->ct_metadata.cookie = (unsigned long)ct | ctinfo;
187 entry->ct_metadata.orig_dir = dir == IP_CT_DIR_ORIGINAL;
189 act_ct_labels = entry->ct_metadata.labels;
190 ct_labels = nf_ct_labels_find(ct);
192 memcpy(act_ct_labels, ct_labels->bits, NF_CT_LABELS_MAX_SIZE);
194 memset(act_ct_labels, 0, NF_CT_LABELS_MAX_SIZE);
197 static int tcf_ct_flow_table_add_action_nat(struct net *net,
199 enum ip_conntrack_dir dir,
200 struct flow_action *action)
202 const struct nf_conntrack_tuple *tuple = &ct->tuplehash[dir].tuple;
203 struct nf_conntrack_tuple target;
205 if (!(ct->status & IPS_NAT_MASK))
208 nf_ct_invert_tuple(&target, &ct->tuplehash[!dir].tuple);
210 switch (tuple->src.l3num) {
212 tcf_ct_flow_table_add_action_nat_ipv4(tuple, target,
216 tcf_ct_flow_table_add_action_nat_ipv6(tuple, target,
223 switch (nf_ct_protonum(ct)) {
225 tcf_ct_flow_table_add_action_nat_tcp(tuple, target, action);
228 tcf_ct_flow_table_add_action_nat_udp(tuple, target, action);
237 static int tcf_ct_flow_table_fill_actions(struct net *net,
238 struct flow_offload *flow,
239 enum flow_offload_tuple_dir tdir,
240 struct nf_flow_rule *flow_rule)
242 struct flow_action *action = &flow_rule->rule->action;
243 int num_entries = action->num_entries;
244 struct nf_conn *ct = flow->ct;
245 enum ip_conntrack_info ctinfo;
246 enum ip_conntrack_dir dir;
250 case FLOW_OFFLOAD_DIR_ORIGINAL:
251 dir = IP_CT_DIR_ORIGINAL;
252 ctinfo = test_bit(IPS_SEEN_REPLY_BIT, &ct->status) ?
253 IP_CT_ESTABLISHED : IP_CT_NEW;
254 if (ctinfo == IP_CT_ESTABLISHED)
255 set_bit(NF_FLOW_HW_ESTABLISHED, &flow->flags);
257 case FLOW_OFFLOAD_DIR_REPLY:
258 dir = IP_CT_DIR_REPLY;
259 ctinfo = IP_CT_ESTABLISHED_REPLY;
265 err = tcf_ct_flow_table_add_action_nat(net, ct, dir, action);
269 tcf_ct_flow_table_add_action_meta(ct, dir, ctinfo, action);
273 /* Clear filled actions */
274 for (i = num_entries; i < action->num_entries; i++)
275 memset(&action->entries[i], 0, sizeof(action->entries[i]));
276 action->num_entries = num_entries;
281 static bool tcf_ct_flow_is_outdated(const struct flow_offload *flow)
283 return test_bit(IPS_SEEN_REPLY_BIT, &flow->ct->status) &&
284 test_bit(IPS_HW_OFFLOAD_BIT, &flow->ct->status) &&
285 !test_bit(NF_FLOW_HW_PENDING, &flow->flags) &&
286 !test_bit(NF_FLOW_HW_ESTABLISHED, &flow->flags);
289 static struct nf_flowtable_type flowtable_ct = {
290 .gc = tcf_ct_flow_is_outdated,
291 .action = tcf_ct_flow_table_fill_actions,
292 .owner = THIS_MODULE,
295 static int tcf_ct_flow_table_get(struct net *net, struct tcf_ct_params *params)
297 struct tcf_ct_flow_table *ct_ft;
300 mutex_lock(&zones_mutex);
301 ct_ft = rhashtable_lookup_fast(&zones_ht, ¶ms->zone, zones_params);
302 if (ct_ft && refcount_inc_not_zero(&ct_ft->ref))
305 ct_ft = kzalloc(sizeof(*ct_ft), GFP_KERNEL);
308 refcount_set(&ct_ft->ref, 1);
310 ct_ft->zone = params->zone;
311 err = rhashtable_insert_fast(&zones_ht, &ct_ft->node, zones_params);
315 ct_ft->nf_ft.type = &flowtable_ct;
316 ct_ft->nf_ft.flags |= NF_FLOWTABLE_HW_OFFLOAD |
317 NF_FLOWTABLE_COUNTER;
318 err = nf_flow_table_init(&ct_ft->nf_ft);
321 write_pnet(&ct_ft->nf_ft.net, net);
323 __module_get(THIS_MODULE);
325 params->ct_ft = ct_ft;
326 params->nf_ft = &ct_ft->nf_ft;
327 mutex_unlock(&zones_mutex);
332 rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params);
336 mutex_unlock(&zones_mutex);
340 static void tcf_ct_flow_table_cleanup_work(struct work_struct *work)
342 struct flow_block_cb *block_cb, *tmp_cb;
343 struct tcf_ct_flow_table *ct_ft;
344 struct flow_block *block;
346 ct_ft = container_of(to_rcu_work(work), struct tcf_ct_flow_table,
348 nf_flow_table_free(&ct_ft->nf_ft);
350 /* Remove any remaining callbacks before cleanup */
351 block = &ct_ft->nf_ft.flow_block;
352 down_write(&ct_ft->nf_ft.flow_block_lock);
353 list_for_each_entry_safe(block_cb, tmp_cb, &block->cb_list, list) {
354 list_del(&block_cb->list);
355 flow_block_cb_free(block_cb);
357 up_write(&ct_ft->nf_ft.flow_block_lock);
360 module_put(THIS_MODULE);
363 static void tcf_ct_flow_table_put(struct tcf_ct_flow_table *ct_ft)
365 if (refcount_dec_and_test(&ct_ft->ref)) {
366 rhashtable_remove_fast(&zones_ht, &ct_ft->node, zones_params);
367 INIT_RCU_WORK(&ct_ft->rwork, tcf_ct_flow_table_cleanup_work);
368 queue_rcu_work(act_ct_wq, &ct_ft->rwork);
372 static void tcf_ct_flow_tc_ifidx(struct flow_offload *entry,
373 struct nf_conn_act_ct_ext *act_ct_ext, u8 dir)
375 entry->tuplehash[dir].tuple.xmit_type = FLOW_OFFLOAD_XMIT_TC;
376 entry->tuplehash[dir].tuple.tc.iifidx = act_ct_ext->ifindex[dir];
379 static void tcf_ct_flow_ct_ext_ifidx_update(struct flow_offload *entry)
381 struct nf_conn_act_ct_ext *act_ct_ext;
383 act_ct_ext = nf_conn_act_ct_ext_find(entry->ct);
385 tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_ORIGINAL);
386 tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_REPLY);
390 static void tcf_ct_flow_table_add(struct tcf_ct_flow_table *ct_ft,
392 bool tcp, bool bidirectional)
394 struct nf_conn_act_ct_ext *act_ct_ext;
395 struct flow_offload *entry;
398 if (test_and_set_bit(IPS_OFFLOAD_BIT, &ct->status))
401 entry = flow_offload_alloc(ct);
408 ct->proto.tcp.seen[0].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
409 ct->proto.tcp.seen[1].flags |= IP_CT_TCP_FLAG_BE_LIBERAL;
412 __set_bit(NF_FLOW_HW_BIDIRECTIONAL, &entry->flags);
414 act_ct_ext = nf_conn_act_ct_ext_find(ct);
416 tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_ORIGINAL);
417 tcf_ct_flow_tc_ifidx(entry, act_ct_ext, FLOW_OFFLOAD_DIR_REPLY);
420 err = flow_offload_add(&ct_ft->nf_ft, entry);
427 flow_offload_free(entry);
429 clear_bit(IPS_OFFLOAD_BIT, &ct->status);
432 static void tcf_ct_flow_table_process_conn(struct tcf_ct_flow_table *ct_ft,
434 enum ip_conntrack_info ctinfo)
436 bool tcp = false, bidirectional = true;
438 switch (nf_ct_protonum(ct)) {
440 if ((ctinfo != IP_CT_ESTABLISHED &&
441 ctinfo != IP_CT_ESTABLISHED_REPLY) ||
442 !test_bit(IPS_ASSURED_BIT, &ct->status) ||
443 ct->proto.tcp.state != TCP_CONNTRACK_ESTABLISHED)
449 if (!nf_ct_is_confirmed(ct))
451 if (!test_bit(IPS_ASSURED_BIT, &ct->status))
452 bidirectional = false;
454 #ifdef CONFIG_NF_CT_PROTO_GRE
456 struct nf_conntrack_tuple *tuple;
458 if ((ctinfo != IP_CT_ESTABLISHED &&
459 ctinfo != IP_CT_ESTABLISHED_REPLY) ||
460 !test_bit(IPS_ASSURED_BIT, &ct->status) ||
461 ct->status & IPS_NAT_MASK)
464 tuple = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
465 /* No support for GRE v1 */
466 if (tuple->src.u.gre.key || tuple->dst.u.gre.key)
475 if (nf_ct_ext_exist(ct, NF_CT_EXT_HELPER) ||
476 ct->status & IPS_SEQ_ADJUST)
479 tcf_ct_flow_table_add(ct_ft, ct, tcp, bidirectional);
483 tcf_ct_flow_table_fill_tuple_ipv4(struct sk_buff *skb,
484 struct flow_offload_tuple *tuple,
485 struct tcphdr **tcph)
487 struct flow_ports *ports;
493 if (!pskb_network_may_pull(skb, sizeof(*iph)))
497 thoff = iph->ihl * 4;
499 if (ip_is_fragment(iph) ||
500 unlikely(thoff != sizeof(struct iphdr)))
503 ipproto = iph->protocol;
506 hdrsize = sizeof(struct tcphdr);
509 hdrsize = sizeof(*ports);
511 #ifdef CONFIG_NF_CT_PROTO_GRE
513 hdrsize = sizeof(struct gre_base_hdr);
523 if (!pskb_network_may_pull(skb, thoff + hdrsize))
528 *tcph = (void *)(skb_network_header(skb) + thoff);
531 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
532 tuple->src_port = ports->source;
533 tuple->dst_port = ports->dest;
536 struct gre_base_hdr *greh;
538 greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
539 if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
547 tuple->src_v4.s_addr = iph->saddr;
548 tuple->dst_v4.s_addr = iph->daddr;
549 tuple->l3proto = AF_INET;
550 tuple->l4proto = ipproto;
556 tcf_ct_flow_table_fill_tuple_ipv6(struct sk_buff *skb,
557 struct flow_offload_tuple *tuple,
558 struct tcphdr **tcph)
560 struct flow_ports *ports;
561 struct ipv6hdr *ip6h;
566 if (!pskb_network_may_pull(skb, sizeof(*ip6h)))
569 ip6h = ipv6_hdr(skb);
570 thoff = sizeof(*ip6h);
572 nexthdr = ip6h->nexthdr;
575 hdrsize = sizeof(struct tcphdr);
578 hdrsize = sizeof(*ports);
580 #ifdef CONFIG_NF_CT_PROTO_GRE
582 hdrsize = sizeof(struct gre_base_hdr);
589 if (ip6h->hop_limit <= 1)
592 if (!pskb_network_may_pull(skb, thoff + hdrsize))
597 *tcph = (void *)(skb_network_header(skb) + thoff);
600 ports = (struct flow_ports *)(skb_network_header(skb) + thoff);
601 tuple->src_port = ports->source;
602 tuple->dst_port = ports->dest;
605 struct gre_base_hdr *greh;
607 greh = (struct gre_base_hdr *)(skb_network_header(skb) + thoff);
608 if ((greh->flags & GRE_VERSION) != GRE_VERSION_0)
614 ip6h = ipv6_hdr(skb);
616 tuple->src_v6 = ip6h->saddr;
617 tuple->dst_v6 = ip6h->daddr;
618 tuple->l3proto = AF_INET6;
619 tuple->l4proto = nexthdr;
624 static bool tcf_ct_flow_table_lookup(struct tcf_ct_params *p,
628 struct nf_flowtable *nf_ft = &p->ct_ft->nf_ft;
629 struct flow_offload_tuple_rhash *tuplehash;
630 struct flow_offload_tuple tuple = {};
631 enum ip_conntrack_info ctinfo;
632 struct tcphdr *tcph = NULL;
633 bool force_refresh = false;
634 struct flow_offload *flow;
640 if (!tcf_ct_flow_table_fill_tuple_ipv4(skb, &tuple, &tcph))
644 if (!tcf_ct_flow_table_fill_tuple_ipv6(skb, &tuple, &tcph))
651 tuplehash = flow_offload_lookup(nf_ft, &tuple);
655 dir = tuplehash->tuple.dir;
656 flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
659 if (dir == FLOW_OFFLOAD_DIR_REPLY &&
660 !test_bit(NF_FLOW_HW_BIDIRECTIONAL, &flow->flags)) {
661 /* Only offload reply direction after connection became
664 if (test_bit(IPS_ASSURED_BIT, &ct->status))
665 set_bit(NF_FLOW_HW_BIDIRECTIONAL, &flow->flags);
666 else if (test_bit(NF_FLOW_HW_ESTABLISHED, &flow->flags))
667 /* If flow_table flow has already been updated to the
668 * established state, then don't refresh.
671 force_refresh = true;
674 if (tcph && (unlikely(tcph->fin || tcph->rst))) {
675 flow_offload_teardown(flow);
679 if (dir == FLOW_OFFLOAD_DIR_ORIGINAL)
680 ctinfo = test_bit(IPS_SEEN_REPLY_BIT, &ct->status) ?
681 IP_CT_ESTABLISHED : IP_CT_NEW;
683 ctinfo = IP_CT_ESTABLISHED_REPLY;
685 nf_conn_act_ct_ext_fill(skb, ct, ctinfo);
686 tcf_ct_flow_ct_ext_ifidx_update(flow);
687 flow_offload_refresh(nf_ft, flow, force_refresh);
688 if (!test_bit(IPS_ASSURED_BIT, &ct->status)) {
689 /* Process this flow in SW to allow promoting to ASSURED */
693 nf_conntrack_get(&ct->ct_general);
694 nf_ct_set(skb, ct, ctinfo);
695 if (nf_ft->flags & NF_FLOWTABLE_COUNTER)
696 nf_ct_acct_update(ct, dir, skb->len);
701 static int tcf_ct_flow_tables_init(void)
703 return rhashtable_init(&zones_ht, &zones_params);
706 static void tcf_ct_flow_tables_uninit(void)
708 rhashtable_destroy(&zones_ht);
711 static struct tc_action_ops act_ct_ops;
713 struct tc_ct_action_net {
714 struct tc_action_net tn; /* Must be first */
717 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
718 static bool tcf_ct_skb_nfct_cached(struct net *net, struct sk_buff *skb,
719 struct tcf_ct_params *p)
721 enum ip_conntrack_info ctinfo;
724 ct = nf_ct_get(skb, &ctinfo);
727 if (!net_eq(net, read_pnet(&ct->ct_net)))
729 if (nf_ct_zone(ct)->id != p->zone)
732 struct nf_conn_help *help;
734 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
735 if (help && rcu_access_pointer(help->helper) != p->helper)
739 /* Force conntrack entry direction. */
740 if ((p->ct_action & TCA_CT_ACT_FORCE) &&
741 CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
742 if (nf_ct_is_confirmed(ct))
752 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
757 static u8 tcf_ct_skb_nf_family(struct sk_buff *skb)
759 u8 family = NFPROTO_UNSPEC;
761 switch (skb_protocol(skb, true)) {
762 case htons(ETH_P_IP):
763 family = NFPROTO_IPV4;
765 case htons(ETH_P_IPV6):
766 family = NFPROTO_IPV6;
775 static int tcf_ct_ipv4_is_fragment(struct sk_buff *skb, bool *frag)
779 len = skb_network_offset(skb) + sizeof(struct iphdr);
780 if (unlikely(skb->len < len))
782 if (unlikely(!pskb_may_pull(skb, len)))
785 *frag = ip_is_fragment(ip_hdr(skb));
789 static int tcf_ct_ipv6_is_fragment(struct sk_buff *skb, bool *frag)
791 unsigned int flags = 0, len, payload_ofs = 0;
792 unsigned short frag_off;
795 len = skb_network_offset(skb) + sizeof(struct ipv6hdr);
796 if (unlikely(skb->len < len))
798 if (unlikely(!pskb_may_pull(skb, len)))
801 nexthdr = ipv6_find_hdr(skb, &payload_ofs, -1, &frag_off, &flags);
802 if (unlikely(nexthdr < 0))
805 *frag = flags & IP6_FH_F_FRAG;
809 static int tcf_ct_handle_fragments(struct net *net, struct sk_buff *skb,
810 u8 family, u16 zone, bool *defrag)
812 enum ip_conntrack_info ctinfo;
819 /* Previously seen (loopback)? Ignore. */
820 ct = nf_ct_get(skb, &ctinfo);
821 if ((ct && !nf_ct_is_template(ct)) || ctinfo == IP_CT_UNTRACKED)
824 if (family == NFPROTO_IPV4)
825 err = tcf_ct_ipv4_is_fragment(skb, &frag);
827 err = tcf_ct_ipv6_is_fragment(skb, &frag);
832 err = nf_ct_handle_fragments(net, skb, zone, family, &proto, &mru);
837 tc_skb_cb(skb)->mru = mru;
842 static void tcf_ct_params_free(struct tcf_ct_params *params)
844 if (params->helper) {
845 #if IS_ENABLED(CONFIG_NF_NAT)
846 if (params->ct_action & TCA_CT_ACT_NAT)
847 nf_nat_helper_put(params->helper);
849 nf_conntrack_helper_put(params->helper);
852 tcf_ct_flow_table_put(params->ct_ft);
854 if (params->put_labels)
855 nf_connlabels_put(nf_ct_net(params->tmpl));
857 nf_ct_put(params->tmpl);
863 static void tcf_ct_params_free_rcu(struct rcu_head *head)
865 struct tcf_ct_params *params;
867 params = container_of(head, struct tcf_ct_params, rcu);
868 tcf_ct_params_free(params);
871 static void tcf_ct_act_set_mark(struct nf_conn *ct, u32 mark, u32 mask)
873 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
879 new_mark = mark | (READ_ONCE(ct->mark) & ~(mask));
880 if (READ_ONCE(ct->mark) != new_mark) {
881 WRITE_ONCE(ct->mark, new_mark);
882 if (nf_ct_is_confirmed(ct))
883 nf_conntrack_event_cache(IPCT_MARK, ct);
888 static void tcf_ct_act_set_labels(struct nf_conn *ct,
892 #if IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)
893 size_t labels_sz = sizeof_field(struct tcf_ct_params, labels);
895 if (!memchr_inv(labels_m, 0, labels_sz))
898 nf_connlabels_replace(ct, labels, labels_m, 4);
902 static int tcf_ct_act_nat(struct sk_buff *skb,
904 enum ip_conntrack_info ctinfo,
906 struct nf_nat_range2 *range,
909 #if IS_ENABLED(CONFIG_NF_NAT)
912 if (!(ct_action & TCA_CT_ACT_NAT))
914 if (ct_action & TCA_CT_ACT_NAT_SRC)
915 action |= BIT(NF_NAT_MANIP_SRC);
916 if (ct_action & TCA_CT_ACT_NAT_DST)
917 action |= BIT(NF_NAT_MANIP_DST);
919 err = nf_ct_nat(skb, ct, ctinfo, &action, range, commit);
921 if (action & BIT(NF_NAT_MANIP_SRC))
922 tc_skb_cb(skb)->post_ct_snat = 1;
923 if (action & BIT(NF_NAT_MANIP_DST))
924 tc_skb_cb(skb)->post_ct_dnat = 1;
932 TC_INDIRECT_SCOPE int tcf_ct_act(struct sk_buff *skb, const struct tc_action *a,
933 struct tcf_result *res)
935 struct net *net = dev_net(skb->dev);
936 enum ip_conntrack_info ctinfo;
937 struct tcf_ct *c = to_ct(a);
938 struct nf_conn *tmpl = NULL;
939 struct nf_hook_state state;
940 bool cached, commit, clear;
941 int nh_ofs, err, retval;
942 struct tcf_ct_params *p;
943 bool add_helper = false;
944 bool skip_add = false;
949 p = rcu_dereference_bh(c->params);
951 retval = READ_ONCE(c->tcf_action);
952 commit = p->ct_action & TCA_CT_ACT_COMMIT;
953 clear = p->ct_action & TCA_CT_ACT_CLEAR;
956 tcf_lastuse_update(&c->tcf_tm);
957 tcf_action_update_bstats(&c->common, skb);
960 tc_skb_cb(skb)->post_ct = false;
961 ct = nf_ct_get(skb, &ctinfo);
964 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
970 family = tcf_ct_skb_nf_family(skb);
971 if (family == NFPROTO_UNSPEC)
974 /* The conntrack module expects to be working at L3.
975 * We also try to pull the IPv4/6 header to linear area
977 nh_ofs = skb_network_offset(skb);
978 skb_pull_rcsum(skb, nh_ofs);
979 err = tcf_ct_handle_fragments(net, skb, family, p->zone, &defrag);
980 if (err == -EINPROGRESS) {
981 retval = TC_ACT_STOLEN;
987 err = nf_ct_skb_network_trim(skb, family);
991 /* If we are recirculating packets to match on ct fields and
992 * committing with a separate ct action, then we don't need to
993 * actually run the packet through conntrack twice unless it's for a
996 cached = tcf_ct_skb_nfct_cached(net, skb, p);
998 if (tcf_ct_flow_table_lookup(p, skb, family)) {
1003 /* Associate skb with specified zone. */
1005 nf_conntrack_put(skb_nfct(skb));
1006 nf_conntrack_get(&tmpl->ct_general);
1007 nf_ct_set(skb, tmpl, IP_CT_NEW);
1010 state.hook = NF_INET_PRE_ROUTING;
1013 err = nf_conntrack_in(skb, &state);
1014 if (err != NF_ACCEPT)
1019 ct = nf_ct_get(skb, &ctinfo);
1022 nf_ct_deliver_cached_events(ct);
1023 nf_conn_act_ct_ext_fill(skb, ct, ctinfo);
1025 err = tcf_ct_act_nat(skb, ct, ctinfo, p->ct_action, &p->range, commit);
1026 if (err != NF_ACCEPT)
1029 if (!nf_ct_is_confirmed(ct) && commit && p->helper && !nfct_help(ct)) {
1030 err = __nf_ct_try_assign_helper(ct, p->tmpl, GFP_ATOMIC);
1034 if (p->ct_action & TCA_CT_ACT_NAT && !nfct_seqadj(ct)) {
1035 if (!nfct_seqadj_ext_add(ct))
1040 if (nf_ct_is_confirmed(ct) ? ((!cached && !skip_add) || add_helper) : commit) {
1041 if (nf_ct_helper(skb, ct, ctinfo, family) != NF_ACCEPT)
1046 tcf_ct_act_set_mark(ct, p->mark, p->mark_mask);
1047 tcf_ct_act_set_labels(ct, p->labels, p->labels_mask);
1049 if (!nf_ct_is_confirmed(ct))
1050 nf_conn_act_ct_ext_add(skb, ct, ctinfo);
1052 /* This will take care of sending queued events
1053 * even if the connection is already confirmed.
1055 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
1060 tcf_ct_flow_table_process_conn(p->ct_ft, ct, ctinfo);
1063 skb_push_rcsum(skb, nh_ofs);
1065 tc_skb_cb(skb)->post_ct = true;
1066 tc_skb_cb(skb)->zone = p->zone;
1069 qdisc_skb_cb(skb)->pkt_len = skb->len;
1073 tcf_action_inc_drop_qstats(&c->common);
1077 static const struct nla_policy ct_policy[TCA_CT_MAX + 1] = {
1078 [TCA_CT_ACTION] = { .type = NLA_U16 },
1079 [TCA_CT_PARMS] = NLA_POLICY_EXACT_LEN(sizeof(struct tc_ct)),
1080 [TCA_CT_ZONE] = { .type = NLA_U16 },
1081 [TCA_CT_MARK] = { .type = NLA_U32 },
1082 [TCA_CT_MARK_MASK] = { .type = NLA_U32 },
1083 [TCA_CT_LABELS] = { .type = NLA_BINARY,
1084 .len = 128 / BITS_PER_BYTE },
1085 [TCA_CT_LABELS_MASK] = { .type = NLA_BINARY,
1086 .len = 128 / BITS_PER_BYTE },
1087 [TCA_CT_NAT_IPV4_MIN] = { .type = NLA_U32 },
1088 [TCA_CT_NAT_IPV4_MAX] = { .type = NLA_U32 },
1089 [TCA_CT_NAT_IPV6_MIN] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
1090 [TCA_CT_NAT_IPV6_MAX] = NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
1091 [TCA_CT_NAT_PORT_MIN] = { .type = NLA_U16 },
1092 [TCA_CT_NAT_PORT_MAX] = { .type = NLA_U16 },
1093 [TCA_CT_HELPER_NAME] = { .type = NLA_STRING, .len = NF_CT_HELPER_NAME_LEN },
1094 [TCA_CT_HELPER_FAMILY] = { .type = NLA_U8 },
1095 [TCA_CT_HELPER_PROTO] = { .type = NLA_U8 },
1098 static int tcf_ct_fill_params_nat(struct tcf_ct_params *p,
1101 struct netlink_ext_ack *extack)
1103 struct nf_nat_range2 *range;
1105 if (!(p->ct_action & TCA_CT_ACT_NAT))
1108 if (!IS_ENABLED(CONFIG_NF_NAT)) {
1109 NL_SET_ERR_MSG_MOD(extack, "Netfilter nat isn't enabled in kernel");
1113 if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
1116 if ((p->ct_action & TCA_CT_ACT_NAT_SRC) &&
1117 (p->ct_action & TCA_CT_ACT_NAT_DST)) {
1118 NL_SET_ERR_MSG_MOD(extack, "dnat and snat can't be enabled at the same time");
1123 if (tb[TCA_CT_NAT_IPV4_MIN]) {
1124 struct nlattr *max_attr = tb[TCA_CT_NAT_IPV4_MAX];
1126 p->ipv4_range = true;
1127 range->flags |= NF_NAT_RANGE_MAP_IPS;
1128 range->min_addr.ip =
1129 nla_get_in_addr(tb[TCA_CT_NAT_IPV4_MIN]);
1131 range->max_addr.ip = max_attr ?
1132 nla_get_in_addr(max_attr) :
1134 } else if (tb[TCA_CT_NAT_IPV6_MIN]) {
1135 struct nlattr *max_attr = tb[TCA_CT_NAT_IPV6_MAX];
1137 p->ipv4_range = false;
1138 range->flags |= NF_NAT_RANGE_MAP_IPS;
1139 range->min_addr.in6 =
1140 nla_get_in6_addr(tb[TCA_CT_NAT_IPV6_MIN]);
1142 range->max_addr.in6 = max_attr ?
1143 nla_get_in6_addr(max_attr) :
1144 range->min_addr.in6;
1147 if (tb[TCA_CT_NAT_PORT_MIN]) {
1148 range->flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1149 range->min_proto.all = nla_get_be16(tb[TCA_CT_NAT_PORT_MIN]);
1151 range->max_proto.all = tb[TCA_CT_NAT_PORT_MAX] ?
1152 nla_get_be16(tb[TCA_CT_NAT_PORT_MAX]) :
1153 range->min_proto.all;
1159 static void tcf_ct_set_key_val(struct nlattr **tb,
1160 void *val, int val_type,
1161 void *mask, int mask_type,
1166 nla_memcpy(val, tb[val_type], len);
1171 if (mask_type == TCA_CT_UNSPEC || !tb[mask_type])
1172 memset(mask, 0xff, len);
1174 nla_memcpy(mask, tb[mask_type], len);
1177 static int tcf_ct_fill_params(struct net *net,
1178 struct tcf_ct_params *p,
1181 struct netlink_ext_ack *extack)
1183 struct nf_conntrack_zone zone;
1184 int err, family, proto, len;
1185 bool put_labels = false;
1186 struct nf_conn *tmpl;
1189 p->zone = NF_CT_DEFAULT_ZONE_ID;
1191 tcf_ct_set_key_val(tb,
1192 &p->ct_action, TCA_CT_ACTION,
1193 NULL, TCA_CT_UNSPEC,
1194 sizeof(p->ct_action));
1196 if (p->ct_action & TCA_CT_ACT_CLEAR)
1199 err = tcf_ct_fill_params_nat(p, parm, tb, extack);
1203 if (tb[TCA_CT_MARK]) {
1204 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)) {
1205 NL_SET_ERR_MSG_MOD(extack, "Conntrack mark isn't enabled.");
1208 tcf_ct_set_key_val(tb,
1209 &p->mark, TCA_CT_MARK,
1210 &p->mark_mask, TCA_CT_MARK_MASK,
1214 if (tb[TCA_CT_LABELS]) {
1215 unsigned int n_bits = sizeof_field(struct tcf_ct_params, labels) * 8;
1217 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS)) {
1218 NL_SET_ERR_MSG_MOD(extack, "Conntrack labels isn't enabled.");
1222 if (nf_connlabels_get(net, n_bits - 1)) {
1223 NL_SET_ERR_MSG_MOD(extack, "Failed to set connlabel length");
1229 tcf_ct_set_key_val(tb,
1230 p->labels, TCA_CT_LABELS,
1231 p->labels_mask, TCA_CT_LABELS_MASK,
1235 if (tb[TCA_CT_ZONE]) {
1236 if (!IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES)) {
1237 NL_SET_ERR_MSG_MOD(extack, "Conntrack zones isn't enabled.");
1241 tcf_ct_set_key_val(tb,
1242 &p->zone, TCA_CT_ZONE,
1243 NULL, TCA_CT_UNSPEC,
1247 nf_ct_zone_init(&zone, p->zone, NF_CT_DEFAULT_ZONE_DIR, 0);
1248 tmpl = nf_ct_tmpl_alloc(net, &zone, GFP_KERNEL);
1250 NL_SET_ERR_MSG_MOD(extack, "Failed to allocate conntrack template");
1254 if (tb[TCA_CT_HELPER_NAME]) {
1255 name = nla_data(tb[TCA_CT_HELPER_NAME]);
1256 len = nla_len(tb[TCA_CT_HELPER_NAME]);
1257 if (len > 16 || name[len - 1] != '\0') {
1258 NL_SET_ERR_MSG_MOD(extack, "Failed to parse helper name.");
1262 family = tb[TCA_CT_HELPER_FAMILY] ? nla_get_u8(tb[TCA_CT_HELPER_FAMILY]) : AF_INET;
1263 proto = tb[TCA_CT_HELPER_PROTO] ? nla_get_u8(tb[TCA_CT_HELPER_PROTO]) : IPPROTO_TCP;
1264 err = nf_ct_add_helper(tmpl, name, family, proto,
1265 p->ct_action & TCA_CT_ACT_NAT, &p->helper);
1267 NL_SET_ERR_MSG_MOD(extack, "Failed to add helper");
1272 p->put_labels = put_labels;
1274 if (p->ct_action & TCA_CT_ACT_COMMIT)
1275 __set_bit(IPS_CONFIRMED_BIT, &tmpl->status);
1279 nf_connlabels_put(net);
1286 static int tcf_ct_init(struct net *net, struct nlattr *nla,
1287 struct nlattr *est, struct tc_action **a,
1288 struct tcf_proto *tp, u32 flags,
1289 struct netlink_ext_ack *extack)
1291 struct tc_action_net *tn = net_generic(net, act_ct_ops.net_id);
1292 bool bind = flags & TCA_ACT_FLAGS_BIND;
1293 struct tcf_ct_params *params = NULL;
1294 struct nlattr *tb[TCA_CT_MAX + 1];
1295 struct tcf_chain *goto_ch = NULL;
1302 NL_SET_ERR_MSG_MOD(extack, "Ct requires attributes to be passed");
1306 err = nla_parse_nested(tb, TCA_CT_MAX, nla, ct_policy, extack);
1310 if (!tb[TCA_CT_PARMS]) {
1311 NL_SET_ERR_MSG_MOD(extack, "Missing required ct parameters");
1314 parm = nla_data(tb[TCA_CT_PARMS]);
1315 index = parm->index;
1316 err = tcf_idr_check_alloc(tn, &index, a, bind);
1321 err = tcf_idr_create_from_flags(tn, index, est, a,
1322 &act_ct_ops, bind, flags);
1324 tcf_idr_cleanup(tn, index);
1327 res = ACT_P_CREATED;
1332 if (!(flags & TCA_ACT_FLAGS_REPLACE)) {
1333 tcf_idr_release(*a, bind);
1337 err = tcf_action_check_ctrlact(parm->action, tp, &goto_ch, extack);
1343 params = kzalloc(sizeof(*params), GFP_KERNEL);
1344 if (unlikely(!params)) {
1349 err = tcf_ct_fill_params(net, params, parm, tb, extack);
1353 err = tcf_ct_flow_table_get(net, params);
1357 spin_lock_bh(&c->tcf_lock);
1358 goto_ch = tcf_action_set_ctrlact(*a, parm->action, goto_ch);
1359 params = rcu_replace_pointer(c->params, params,
1360 lockdep_is_held(&c->tcf_lock));
1361 spin_unlock_bh(&c->tcf_lock);
1364 tcf_chain_put_by_act(goto_ch);
1366 call_rcu(¶ms->rcu, tcf_ct_params_free_rcu);
1372 tcf_chain_put_by_act(goto_ch);
1374 tcf_ct_params_free(params);
1375 tcf_idr_release(*a, bind);
1379 static void tcf_ct_cleanup(struct tc_action *a)
1381 struct tcf_ct_params *params;
1382 struct tcf_ct *c = to_ct(a);
1384 params = rcu_dereference_protected(c->params, 1);
1386 call_rcu(¶ms->rcu, tcf_ct_params_free_rcu);
1389 static int tcf_ct_dump_key_val(struct sk_buff *skb,
1390 void *val, int val_type,
1391 void *mask, int mask_type,
1396 if (mask && !memchr_inv(mask, 0, len))
1399 err = nla_put(skb, val_type, len, val);
1403 if (mask_type != TCA_CT_UNSPEC) {
1404 err = nla_put(skb, mask_type, len, mask);
1412 static int tcf_ct_dump_nat(struct sk_buff *skb, struct tcf_ct_params *p)
1414 struct nf_nat_range2 *range = &p->range;
1416 if (!(p->ct_action & TCA_CT_ACT_NAT))
1419 if (!(p->ct_action & (TCA_CT_ACT_NAT_SRC | TCA_CT_ACT_NAT_DST)))
1422 if (range->flags & NF_NAT_RANGE_MAP_IPS) {
1423 if (p->ipv4_range) {
1424 if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MIN,
1425 range->min_addr.ip))
1427 if (nla_put_in_addr(skb, TCA_CT_NAT_IPV4_MAX,
1428 range->max_addr.ip))
1431 if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MIN,
1432 &range->min_addr.in6))
1434 if (nla_put_in6_addr(skb, TCA_CT_NAT_IPV6_MAX,
1435 &range->max_addr.in6))
1440 if (range->flags & NF_NAT_RANGE_PROTO_SPECIFIED) {
1441 if (nla_put_be16(skb, TCA_CT_NAT_PORT_MIN,
1442 range->min_proto.all))
1444 if (nla_put_be16(skb, TCA_CT_NAT_PORT_MAX,
1445 range->max_proto.all))
1452 static int tcf_ct_dump_helper(struct sk_buff *skb, struct nf_conntrack_helper *helper)
1457 if (nla_put_string(skb, TCA_CT_HELPER_NAME, helper->name) ||
1458 nla_put_u8(skb, TCA_CT_HELPER_FAMILY, helper->tuple.src.l3num) ||
1459 nla_put_u8(skb, TCA_CT_HELPER_PROTO, helper->tuple.dst.protonum))
1465 static inline int tcf_ct_dump(struct sk_buff *skb, struct tc_action *a,
1468 unsigned char *b = skb_tail_pointer(skb);
1469 struct tcf_ct *c = to_ct(a);
1470 struct tcf_ct_params *p;
1472 struct tc_ct opt = {
1473 .index = c->tcf_index,
1474 .refcnt = refcount_read(&c->tcf_refcnt) - ref,
1475 .bindcnt = atomic_read(&c->tcf_bindcnt) - bind,
1479 spin_lock_bh(&c->tcf_lock);
1480 p = rcu_dereference_protected(c->params,
1481 lockdep_is_held(&c->tcf_lock));
1482 opt.action = c->tcf_action;
1484 if (tcf_ct_dump_key_val(skb,
1485 &p->ct_action, TCA_CT_ACTION,
1486 NULL, TCA_CT_UNSPEC,
1487 sizeof(p->ct_action)))
1488 goto nla_put_failure;
1490 if (p->ct_action & TCA_CT_ACT_CLEAR)
1493 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1494 tcf_ct_dump_key_val(skb,
1495 &p->mark, TCA_CT_MARK,
1496 &p->mark_mask, TCA_CT_MARK_MASK,
1498 goto nla_put_failure;
1500 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1501 tcf_ct_dump_key_val(skb,
1502 p->labels, TCA_CT_LABELS,
1503 p->labels_mask, TCA_CT_LABELS_MASK,
1505 goto nla_put_failure;
1507 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1508 tcf_ct_dump_key_val(skb,
1509 &p->zone, TCA_CT_ZONE,
1510 NULL, TCA_CT_UNSPEC,
1512 goto nla_put_failure;
1514 if (tcf_ct_dump_nat(skb, p))
1515 goto nla_put_failure;
1517 if (tcf_ct_dump_helper(skb, p->helper))
1518 goto nla_put_failure;
1521 if (nla_put(skb, TCA_CT_PARMS, sizeof(opt), &opt))
1522 goto nla_put_failure;
1524 tcf_tm_dump(&t, &c->tcf_tm);
1525 if (nla_put_64bit(skb, TCA_CT_TM, sizeof(t), &t, TCA_CT_PAD))
1526 goto nla_put_failure;
1527 spin_unlock_bh(&c->tcf_lock);
1531 spin_unlock_bh(&c->tcf_lock);
1536 static void tcf_stats_update(struct tc_action *a, u64 bytes, u64 packets,
1537 u64 drops, u64 lastuse, bool hw)
1539 struct tcf_ct *c = to_ct(a);
1541 tcf_action_update_stats(a, bytes, packets, drops, hw);
1542 c->tcf_tm.lastuse = max_t(u64, c->tcf_tm.lastuse, lastuse);
1545 static int tcf_ct_offload_act_setup(struct tc_action *act, void *entry_data,
1546 u32 *index_inc, bool bind,
1547 struct netlink_ext_ack *extack)
1550 struct flow_action_entry *entry = entry_data;
1552 entry->id = FLOW_ACTION_CT;
1553 entry->ct.action = tcf_ct_action(act);
1554 entry->ct.zone = tcf_ct_zone(act);
1555 entry->ct.flow_table = tcf_ct_ft(act);
1558 struct flow_offload_action *fl_action = entry_data;
1560 fl_action->id = FLOW_ACTION_CT;
1566 static struct tc_action_ops act_ct_ops = {
1569 .owner = THIS_MODULE,
1571 .dump = tcf_ct_dump,
1572 .init = tcf_ct_init,
1573 .cleanup = tcf_ct_cleanup,
1574 .stats_update = tcf_stats_update,
1575 .offload_act_setup = tcf_ct_offload_act_setup,
1576 .size = sizeof(struct tcf_ct),
1579 static __net_init int ct_init_net(struct net *net)
1581 struct tc_ct_action_net *tn = net_generic(net, act_ct_ops.net_id);
1583 return tc_action_net_init(net, &tn->tn, &act_ct_ops);
1586 static void __net_exit ct_exit_net(struct list_head *net_list)
1588 tc_action_net_exit(net_list, act_ct_ops.net_id);
1591 static struct pernet_operations ct_net_ops = {
1592 .init = ct_init_net,
1593 .exit_batch = ct_exit_net,
1594 .id = &act_ct_ops.net_id,
1595 .size = sizeof(struct tc_ct_action_net),
1598 static int __init ct_init_module(void)
1602 act_ct_wq = alloc_ordered_workqueue("act_ct_workqueue", 0);
1606 err = tcf_ct_flow_tables_init();
1610 err = tcf_register_action(&act_ct_ops, &ct_net_ops);
1614 static_branch_inc(&tcf_frag_xmit_count);
1619 tcf_ct_flow_tables_uninit();
1621 destroy_workqueue(act_ct_wq);
1625 static void __exit ct_cleanup_module(void)
1627 static_branch_dec(&tcf_frag_xmit_count);
1628 tcf_unregister_action(&act_ct_ops, &ct_net_ops);
1629 tcf_ct_flow_tables_uninit();
1630 destroy_workqueue(act_ct_wq);
1633 module_init(ct_init_module);
1634 module_exit(ct_cleanup_module);
1635 MODULE_AUTHOR("Paul Blakey <paulb@mellanox.com>");
1636 MODULE_AUTHOR("Yossi Kuperman <yossiku@mellanox.com>");
1637 MODULE_AUTHOR("Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>");
1638 MODULE_DESCRIPTION("Connection tracking action");
1639 MODULE_LICENSE("GPL v2");