2 * Copyright (c) 2015 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
14 #include <linux/module.h>
15 #include <linux/openvswitch.h>
16 #include <linux/tcp.h>
17 #include <linux/udp.h>
18 #include <linux/sctp.h>
20 #include <net/netfilter/nf_conntrack_core.h>
21 #include <net/netfilter/nf_conntrack_helper.h>
22 #include <net/netfilter/nf_conntrack_labels.h>
23 #include <net/netfilter/nf_conntrack_seqadj.h>
24 #include <net/netfilter/nf_conntrack_zones.h>
25 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
27 #ifdef CONFIG_NF_NAT_NEEDED
28 #include <linux/netfilter/nf_nat.h>
29 #include <net/netfilter/nf_nat_core.h>
30 #include <net/netfilter/nf_nat_l3proto.h>
34 #include "conntrack.h"
36 #include "flow_netlink.h"
38 struct ovs_ct_len_tbl {
43 /* Metadata mark for masked write to conntrack mark */
49 /* Metadata label for masked write to conntrack label. */
51 struct ovs_key_ct_labels value;
52 struct ovs_key_ct_labels mask;
56 OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */
57 OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */
58 OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */
61 /* Conntrack action context for execution. */
62 struct ovs_conntrack_info {
63 struct nf_conntrack_helper *helper;
64 struct nf_conntrack_zone zone;
67 u8 nat : 3; /* enum ovs_ct_nat */
69 u8 have_eventmask : 1;
71 u32 eventmask; /* Mask of 1 << IPCT_*. */
73 struct md_labels labels;
74 #ifdef CONFIG_NF_NAT_NEEDED
75 struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */
79 static bool labels_nonzero(const struct ovs_key_ct_labels *labels);
81 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
83 static u16 key_to_nfproto(const struct sw_flow_key *key)
85 switch (ntohs(key->eth.type)) {
91 return NFPROTO_UNSPEC;
95 /* Map SKB connection state into the values used by flow definition. */
96 static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
98 u8 ct_state = OVS_CS_F_TRACKED;
101 case IP_CT_ESTABLISHED_REPLY:
102 case IP_CT_RELATED_REPLY:
103 ct_state |= OVS_CS_F_REPLY_DIR;
110 case IP_CT_ESTABLISHED:
111 case IP_CT_ESTABLISHED_REPLY:
112 ct_state |= OVS_CS_F_ESTABLISHED;
115 case IP_CT_RELATED_REPLY:
116 ct_state |= OVS_CS_F_RELATED;
119 ct_state |= OVS_CS_F_NEW;
128 static u32 ovs_ct_get_mark(const struct nf_conn *ct)
130 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
131 return ct ? ct->mark : 0;
137 /* Guard against conntrack labels max size shrinking below 128 bits. */
138 #if NF_CT_LABELS_MAX_SIZE < 16
139 #error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes
142 static void ovs_ct_get_labels(const struct nf_conn *ct,
143 struct ovs_key_ct_labels *labels)
145 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
148 memcpy(labels, cl->bits, OVS_CT_LABELS_LEN);
150 memset(labels, 0, OVS_CT_LABELS_LEN);
153 static void __ovs_ct_update_key_orig_tp(struct sw_flow_key *key,
154 const struct nf_conntrack_tuple *orig,
157 key->ct_orig_proto = orig->dst.protonum;
158 if (orig->dst.protonum == icmp_proto) {
159 key->ct.orig_tp.src = htons(orig->dst.u.icmp.type);
160 key->ct.orig_tp.dst = htons(orig->dst.u.icmp.code);
162 key->ct.orig_tp.src = orig->src.u.all;
163 key->ct.orig_tp.dst = orig->dst.u.all;
167 static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
168 const struct nf_conntrack_zone *zone,
169 const struct nf_conn *ct)
171 key->ct_state = state;
172 key->ct_zone = zone->id;
173 key->ct.mark = ovs_ct_get_mark(ct);
174 ovs_ct_get_labels(ct, &key->ct.labels);
177 const struct nf_conntrack_tuple *orig;
179 /* Use the master if we have one. */
182 orig = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
184 /* IP version must match with the master connection. */
185 if (key->eth.type == htons(ETH_P_IP) &&
186 nf_ct_l3num(ct) == NFPROTO_IPV4) {
187 key->ipv4.ct_orig.src = orig->src.u3.ip;
188 key->ipv4.ct_orig.dst = orig->dst.u3.ip;
189 __ovs_ct_update_key_orig_tp(key, orig, IPPROTO_ICMP);
191 } else if (key->eth.type == htons(ETH_P_IPV6) &&
192 !sw_flow_key_is_nd(key) &&
193 nf_ct_l3num(ct) == NFPROTO_IPV6) {
194 key->ipv6.ct_orig.src = orig->src.u3.in6;
195 key->ipv6.ct_orig.dst = orig->dst.u3.in6;
196 __ovs_ct_update_key_orig_tp(key, orig, NEXTHDR_ICMP);
200 /* Clear 'ct_orig_proto' to mark the non-existence of conntrack
201 * original direction key fields.
203 key->ct_orig_proto = 0;
206 /* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has
207 * previously sent the packet to conntrack via the ct action. If
208 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
209 * initialized from the connection status.
211 static void ovs_ct_update_key(const struct sk_buff *skb,
212 const struct ovs_conntrack_info *info,
213 struct sw_flow_key *key, bool post_ct,
216 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
217 enum ip_conntrack_info ctinfo;
221 ct = nf_ct_get(skb, &ctinfo);
223 state = ovs_ct_get_state(ctinfo);
224 /* All unconfirmed entries are NEW connections. */
225 if (!nf_ct_is_confirmed(ct))
226 state |= OVS_CS_F_NEW;
227 /* OVS persists the related flag for the duration of the
231 state |= OVS_CS_F_RELATED;
232 if (keep_nat_flags) {
233 state |= key->ct_state & OVS_CS_F_NAT_MASK;
235 if (ct->status & IPS_SRC_NAT)
236 state |= OVS_CS_F_SRC_NAT;
237 if (ct->status & IPS_DST_NAT)
238 state |= OVS_CS_F_DST_NAT;
240 zone = nf_ct_zone(ct);
241 } else if (post_ct) {
242 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
246 __ovs_ct_update_key(key, state, zone, ct);
249 /* This is called to initialize CT key fields possibly coming in from the local
252 void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
254 ovs_ct_update_key(skb, NULL, key, false, false);
257 #define IN6_ADDR_INITIALIZER(ADDR) \
258 { (ADDR).s6_addr32[0], (ADDR).s6_addr32[1], \
259 (ADDR).s6_addr32[2], (ADDR).s6_addr32[3] }
261 int ovs_ct_put_key(const struct sw_flow_key *swkey,
262 const struct sw_flow_key *output, struct sk_buff *skb)
264 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, output->ct_state))
267 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
268 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, output->ct_zone))
271 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
272 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, output->ct.mark))
275 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
276 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(output->ct.labels),
280 if (swkey->ct_orig_proto) {
281 if (swkey->eth.type == htons(ETH_P_IP)) {
282 struct ovs_key_ct_tuple_ipv4 orig = {
283 output->ipv4.ct_orig.src,
284 output->ipv4.ct_orig.dst,
285 output->ct.orig_tp.src,
286 output->ct.orig_tp.dst,
287 output->ct_orig_proto,
289 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4,
290 sizeof(orig), &orig))
292 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
293 struct ovs_key_ct_tuple_ipv6 orig = {
294 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.src),
295 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.dst),
296 output->ct.orig_tp.src,
297 output->ct.orig_tp.dst,
298 output->ct_orig_proto,
300 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6,
301 sizeof(orig), &orig))
309 static int ovs_ct_set_mark(struct nf_conn *ct, struct sw_flow_key *key,
310 u32 ct_mark, u32 mask)
312 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
315 new_mark = ct_mark | (ct->mark & ~(mask));
316 if (ct->mark != new_mark) {
318 if (nf_ct_is_confirmed(ct))
319 nf_conntrack_event_cache(IPCT_MARK, ct);
320 key->ct.mark = new_mark;
329 static struct nf_conn_labels *ovs_ct_get_conn_labels(struct nf_conn *ct)
331 struct nf_conn_labels *cl;
333 cl = nf_ct_labels_find(ct);
335 nf_ct_labels_ext_add(ct);
336 cl = nf_ct_labels_find(ct);
342 /* Initialize labels for a new, yet to be committed conntrack entry. Note that
343 * since the new connection is not yet confirmed, and thus no-one else has
344 * access to it's labels, we simply write them over.
346 static int ovs_ct_init_labels(struct nf_conn *ct, struct sw_flow_key *key,
347 const struct ovs_key_ct_labels *labels,
348 const struct ovs_key_ct_labels *mask)
350 struct nf_conn_labels *cl, *master_cl;
351 bool have_mask = labels_nonzero(mask);
353 /* Inherit master's labels to the related connection? */
354 master_cl = ct->master ? nf_ct_labels_find(ct->master) : NULL;
356 if (!master_cl && !have_mask)
357 return 0; /* Nothing to do. */
359 cl = ovs_ct_get_conn_labels(ct);
363 /* Inherit the master's labels, if any. */
368 u32 *dst = (u32 *)cl->bits;
371 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
372 dst[i] = (dst[i] & ~mask->ct_labels_32[i]) |
373 (labels->ct_labels_32[i]
374 & mask->ct_labels_32[i]);
377 /* Labels are included in the IPCTNL_MSG_CT_NEW event only if the
378 * IPCT_LABEL bit is set in the event cache.
380 nf_conntrack_event_cache(IPCT_LABEL, ct);
382 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
387 static int ovs_ct_set_labels(struct nf_conn *ct, struct sw_flow_key *key,
388 const struct ovs_key_ct_labels *labels,
389 const struct ovs_key_ct_labels *mask)
391 struct nf_conn_labels *cl;
394 cl = ovs_ct_get_conn_labels(ct);
398 err = nf_connlabels_replace(ct, labels->ct_labels_32,
400 OVS_CT_LABELS_LEN_32);
404 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
409 /* 'skb' should already be pulled to nh_ofs. */
410 static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
412 const struct nf_conntrack_helper *helper;
413 const struct nf_conn_help *help;
414 enum ip_conntrack_info ctinfo;
415 unsigned int protoff;
419 ct = nf_ct_get(skb, &ctinfo);
420 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
423 help = nfct_help(ct);
427 helper = rcu_dereference(help->helper);
433 protoff = ip_hdrlen(skb);
436 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
440 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
442 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
443 pr_debug("proto header not found\n");
450 WARN_ONCE(1, "helper invoked on non-IP family!");
454 err = helper->help(skb, protoff, ct, ctinfo);
455 if (err != NF_ACCEPT)
458 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
459 * FTP with NAT) adusting the TCP payload size when mangling IP
460 * addresses and/or port numbers in the text-based control connection.
462 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
463 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
468 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
469 * value if 'skb' is freed.
471 static int handle_fragments(struct net *net, struct sw_flow_key *key,
472 u16 zone, struct sk_buff *skb)
474 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
477 if (key->eth.type == htons(ETH_P_IP)) {
478 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
480 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
481 err = ip_defrag(net, skb, user);
485 ovs_cb.mru = IPCB(skb)->frag_max_size;
486 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
487 } else if (key->eth.type == htons(ETH_P_IPV6)) {
488 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
490 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
491 err = nf_ct_frag6_gather(net, skb, user);
493 if (err != -EINPROGRESS)
498 key->ip.proto = ipv6_hdr(skb)->nexthdr;
499 ovs_cb.mru = IP6CB(skb)->frag_max_size;
503 return -EPFNOSUPPORT;
506 key->ip.frag = OVS_FRAG_TYPE_NONE;
509 *OVS_CB(skb) = ovs_cb;
514 static struct nf_conntrack_expect *
515 ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
516 u16 proto, const struct sk_buff *skb)
518 struct nf_conntrack_tuple tuple;
519 struct nf_conntrack_expect *exp;
521 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
524 exp = __nf_ct_expect_find(net, zone, &tuple);
526 struct nf_conntrack_tuple_hash *h;
528 /* Delete existing conntrack entry, if it clashes with the
529 * expectation. This can happen since conntrack ALGs do not
530 * check for clashes between (new) expectations and existing
531 * conntrack entries. nf_conntrack_in() will check the
532 * expectations only if a conntrack entry can not be found,
533 * which can lead to OVS finding the expectation (here) in the
534 * init direction, but which will not be removed by the
535 * nf_conntrack_in() call, if a matching conntrack entry is
536 * found instead. In this case all init direction packets
537 * would be reported as new related packets, while reply
538 * direction packets would be reported as un-related
539 * established packets.
541 h = nf_conntrack_find_get(net, zone, &tuple);
543 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
545 nf_ct_delete(ct, 0, 0);
546 nf_conntrack_put(&ct->ct_general);
553 /* This replicates logic from nf_conntrack_core.c that is not exported. */
554 static enum ip_conntrack_info
555 ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
557 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
559 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
560 return IP_CT_ESTABLISHED_REPLY;
561 /* Once we've had two way comms, always ESTABLISHED. */
562 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
563 return IP_CT_ESTABLISHED;
564 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
565 return IP_CT_RELATED;
569 /* Find an existing connection which this packet belongs to without
570 * re-attributing statistics or modifying the connection state. This allows an
571 * skb->_nfct lost due to an upcall to be recovered during actions execution.
573 * Must be called with rcu_read_lock.
575 * On success, populates skb->_nfct and returns the connection. Returns NULL
576 * if there is no existing entry.
578 static struct nf_conn *
579 ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
580 u8 l3num, struct sk_buff *skb, bool natted)
582 struct nf_conntrack_l3proto *l3proto;
583 struct nf_conntrack_l4proto *l4proto;
584 struct nf_conntrack_tuple tuple;
585 struct nf_conntrack_tuple_hash *h;
587 unsigned int dataoff;
590 l3proto = __nf_ct_l3proto_find(l3num);
591 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
593 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
596 l4proto = __nf_ct_l4proto_find(l3num, protonum);
597 if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
598 protonum, net, &tuple, l3proto, l4proto)) {
599 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
603 /* Must invert the tuple if skb has been transformed by NAT. */
605 struct nf_conntrack_tuple inverse;
607 if (!nf_ct_invert_tuple(&inverse, &tuple, l3proto, l4proto)) {
608 pr_debug("ovs_ct_find_existing: Inversion failed!\n");
614 /* look for tuple match */
615 h = nf_conntrack_find_get(net, zone, &tuple);
617 return NULL; /* Not found. */
619 ct = nf_ct_tuplehash_to_ctrack(h);
621 /* Inverted packet tuple matches the reverse direction conntrack tuple,
622 * select the other tuplehash to get the right 'ctinfo' bits for this
626 h = &ct->tuplehash[!h->tuple.dst.dir];
628 nf_ct_set(skb, ct, ovs_ct_get_info(h));
632 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
633 static bool skb_nfct_cached(struct net *net,
634 const struct sw_flow_key *key,
635 const struct ovs_conntrack_info *info,
638 enum ip_conntrack_info ctinfo;
641 ct = nf_ct_get(skb, &ctinfo);
642 /* If no ct, check if we have evidence that an existing conntrack entry
643 * might be found for this skb. This happens when we lose a skb->_nfct
644 * due to an upcall. If the connection was not confirmed, it is not
645 * cached and needs to be run through conntrack again.
647 if (!ct && key->ct_state & OVS_CS_F_TRACKED &&
648 !(key->ct_state & OVS_CS_F_INVALID) &&
649 key->ct_zone == info->zone.id) {
650 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb,
652 & OVS_CS_F_NAT_MASK));
654 nf_ct_get(skb, &ctinfo);
658 if (!net_eq(net, read_pnet(&ct->ct_net)))
660 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
663 struct nf_conn_help *help;
665 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
666 if (help && rcu_access_pointer(help->helper) != info->helper)
669 /* Force conntrack entry direction to the current packet? */
670 if (info->force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
671 /* Delete the conntrack entry if confirmed, else just release
674 if (nf_ct_is_confirmed(ct))
675 nf_ct_delete(ct, 0, 0);
677 nf_conntrack_put(&ct->ct_general);
678 nf_ct_set(skb, NULL, 0);
685 #ifdef CONFIG_NF_NAT_NEEDED
686 /* Modelled after nf_nat_ipv[46]_fn().
687 * range is only used for new, uninitialized NAT state.
688 * Returns either NF_ACCEPT or NF_DROP.
690 static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
691 enum ip_conntrack_info ctinfo,
692 const struct nf_nat_range *range,
693 enum nf_nat_manip_type maniptype)
695 int hooknum, nh_off, err = NF_ACCEPT;
697 nh_off = skb_network_offset(skb);
698 skb_pull_rcsum(skb, nh_off);
700 /* See HOOK2MANIP(). */
701 if (maniptype == NF_NAT_MANIP_SRC)
702 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
704 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
708 case IP_CT_RELATED_REPLY:
709 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
710 skb->protocol == htons(ETH_P_IP) &&
711 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
712 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
716 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
717 skb->protocol == htons(ETH_P_IPV6)) {
719 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
720 int hdrlen = ipv6_skip_exthdr(skb,
721 sizeof(struct ipv6hdr),
722 &nexthdr, &frag_off);
724 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
725 if (!nf_nat_icmpv6_reply_translation(skb, ct,
733 /* Non-ICMP, fall thru to initialize if needed. */
735 /* Seen it before? This can happen for loopback, retrans,
738 if (!nf_nat_initialized(ct, maniptype)) {
739 /* Initialize according to the NAT action. */
740 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
741 /* Action is set up to establish a new
744 ? nf_nat_setup_info(ct, range, maniptype)
745 : nf_nat_alloc_null_binding(ct, hooknum);
746 if (err != NF_ACCEPT)
751 case IP_CT_ESTABLISHED:
752 case IP_CT_ESTABLISHED_REPLY:
760 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
762 skb_push(skb, nh_off);
763 skb_postpush_rcsum(skb, skb->data, nh_off);
768 static void ovs_nat_update_key(struct sw_flow_key *key,
769 const struct sk_buff *skb,
770 enum nf_nat_manip_type maniptype)
772 if (maniptype == NF_NAT_MANIP_SRC) {
775 key->ct_state |= OVS_CS_F_SRC_NAT;
776 if (key->eth.type == htons(ETH_P_IP))
777 key->ipv4.addr.src = ip_hdr(skb)->saddr;
778 else if (key->eth.type == htons(ETH_P_IPV6))
779 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
780 sizeof(key->ipv6.addr.src));
784 if (key->ip.proto == IPPROTO_UDP)
785 src = udp_hdr(skb)->source;
786 else if (key->ip.proto == IPPROTO_TCP)
787 src = tcp_hdr(skb)->source;
788 else if (key->ip.proto == IPPROTO_SCTP)
789 src = sctp_hdr(skb)->source;
797 key->ct_state |= OVS_CS_F_DST_NAT;
798 if (key->eth.type == htons(ETH_P_IP))
799 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
800 else if (key->eth.type == htons(ETH_P_IPV6))
801 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
802 sizeof(key->ipv6.addr.dst));
806 if (key->ip.proto == IPPROTO_UDP)
807 dst = udp_hdr(skb)->dest;
808 else if (key->ip.proto == IPPROTO_TCP)
809 dst = tcp_hdr(skb)->dest;
810 else if (key->ip.proto == IPPROTO_SCTP)
811 dst = sctp_hdr(skb)->dest;
819 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
820 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
821 const struct ovs_conntrack_info *info,
822 struct sk_buff *skb, struct nf_conn *ct,
823 enum ip_conntrack_info ctinfo)
825 enum nf_nat_manip_type maniptype;
828 /* Add NAT extension if not confirmed yet. */
829 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
830 return NF_ACCEPT; /* Can't NAT. */
832 /* Determine NAT type.
833 * Check if the NAT type can be deduced from the tracked connection.
834 * Make sure new expected connections (IP_CT_RELATED) are NATted only
837 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
838 ct->status & IPS_NAT_MASK &&
839 (ctinfo != IP_CT_RELATED || info->commit)) {
840 /* NAT an established or related connection like before. */
841 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
842 /* This is the REPLY direction for a connection
843 * for which NAT was applied in the forward
844 * direction. Do the reverse NAT.
846 maniptype = ct->status & IPS_SRC_NAT
847 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
849 maniptype = ct->status & IPS_SRC_NAT
850 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
851 } else if (info->nat & OVS_CT_SRC_NAT) {
852 maniptype = NF_NAT_MANIP_SRC;
853 } else if (info->nat & OVS_CT_DST_NAT) {
854 maniptype = NF_NAT_MANIP_DST;
856 return NF_ACCEPT; /* Connection is not NATed. */
858 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
860 /* Mark NAT done if successful and update the flow key. */
861 if (err == NF_ACCEPT)
862 ovs_nat_update_key(key, skb, maniptype);
866 #else /* !CONFIG_NF_NAT_NEEDED */
867 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
868 const struct ovs_conntrack_info *info,
869 struct sk_buff *skb, struct nf_conn *ct,
870 enum ip_conntrack_info ctinfo)
876 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
877 * not done already. Update key with new CT state after passing the packet
879 * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be
880 * set to NULL and 0 will be returned.
882 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
883 const struct ovs_conntrack_info *info,
886 /* If we are recirculating packets to match on conntrack fields and
887 * committing with a separate conntrack action, then we don't need to
888 * actually run the packet through conntrack twice unless it's for a
891 bool cached = skb_nfct_cached(net, key, info, skb);
892 enum ip_conntrack_info ctinfo;
896 struct nf_conn *tmpl = info->ct;
899 /* Associate skb with specified zone. */
902 nf_conntrack_put(skb_nfct(skb));
903 nf_conntrack_get(&tmpl->ct_general);
904 nf_ct_set(skb, tmpl, IP_CT_NEW);
907 err = nf_conntrack_in(net, info->family,
908 NF_INET_PRE_ROUTING, skb);
909 if (err != NF_ACCEPT)
912 /* Clear CT state NAT flags to mark that we have not yet done
913 * NAT after the nf_conntrack_in() call. We can actually clear
914 * the whole state, as it will be re-initialized below.
918 /* Update the key, but keep the NAT flags. */
919 ovs_ct_update_key(skb, info, key, true, true);
922 ct = nf_ct_get(skb, &ctinfo);
924 /* Packets starting a new connection must be NATted before the
925 * helper, so that the helper knows about the NAT. We enforce
926 * this by delaying both NAT and helper calls for unconfirmed
927 * connections until the committing CT action. For later
928 * packets NAT and Helper may be called in either order.
930 * NAT will be done only if the CT action has NAT, and only
931 * once per packet (per zone), as guarded by the NAT bits in
934 if (info->nat && !(key->ct_state & OVS_CS_F_NAT_MASK) &&
935 (nf_ct_is_confirmed(ct) || info->commit) &&
936 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
940 /* Userspace may decide to perform a ct lookup without a helper
941 * specified followed by a (recirculate and) commit with one.
942 * Therefore, for unconfirmed connections which we will commit,
943 * we need to attach the helper here.
945 if (!nf_ct_is_confirmed(ct) && info->commit &&
946 info->helper && !nfct_help(ct)) {
947 int err = __nf_ct_try_assign_helper(ct, info->ct,
953 /* Call the helper only if:
954 * - nf_conntrack_in() was executed above ("!cached") for a
955 * confirmed connection, or
956 * - When committing an unconfirmed connection.
958 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
959 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
967 /* Lookup connection and read fields into key. */
968 static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
969 const struct ovs_conntrack_info *info,
972 struct nf_conntrack_expect *exp;
974 /* If we pass an expected packet through nf_conntrack_in() the
975 * expectation is typically removed, but the packet could still be
976 * lost in upcall processing. To prevent this from happening we
977 * perform an explicit expectation lookup. Expected connections are
978 * always new, and will be passed through conntrack only when they are
979 * committed, as it is OK to remove the expectation at that time.
981 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
985 /* NOTE: New connections are NATted and Helped only when
986 * committed, so we are not calling into NAT here.
988 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
989 __ovs_ct_update_key(key, state, &info->zone, exp->master);
994 err = __ovs_ct_lookup(net, key, info, skb);
998 ct = (struct nf_conn *)skb_nfct(skb);
1000 nf_ct_deliver_cached_events(ct);
1006 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
1010 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
1011 if (labels->ct_labels_32[i])
1017 /* Lookup connection and confirm if unconfirmed. */
1018 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
1019 const struct ovs_conntrack_info *info,
1020 struct sk_buff *skb)
1022 enum ip_conntrack_info ctinfo;
1026 err = __ovs_ct_lookup(net, key, info, skb);
1030 /* The connection could be invalid, in which case this is a no-op.*/
1031 ct = nf_ct_get(skb, &ctinfo);
1035 /* Set the conntrack event mask if given. NEW and DELETE events have
1036 * their own groups, but the NFNLGRP_CONNTRACK_UPDATE group listener
1037 * typically would receive many kinds of updates. Setting the event
1038 * mask allows those events to be filtered. The set event mask will
1039 * remain in effect for the lifetime of the connection unless changed
1040 * by a further CT action with both the commit flag and the eventmask
1042 if (info->have_eventmask) {
1043 struct nf_conntrack_ecache *cache = nf_ct_ecache_find(ct);
1046 cache->ctmask = info->eventmask;
1049 /* Apply changes before confirming the connection so that the initial
1050 * conntrack NEW netlink event carries the values given in the CT
1053 if (info->mark.mask) {
1054 err = ovs_ct_set_mark(ct, key, info->mark.value,
1059 if (!nf_ct_is_confirmed(ct)) {
1060 err = ovs_ct_init_labels(ct, key, &info->labels.value,
1061 &info->labels.mask);
1064 } else if (labels_nonzero(&info->labels.mask)) {
1065 err = ovs_ct_set_labels(ct, key, &info->labels.value,
1066 &info->labels.mask);
1070 /* This will take care of sending queued events even if the connection
1071 * is already confirmed.
1073 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
1079 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
1080 * value if 'skb' is freed.
1082 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
1083 struct sw_flow_key *key,
1084 const struct ovs_conntrack_info *info)
1089 /* The conntrack module expects to be working at L3. */
1090 nh_ofs = skb_network_offset(skb);
1091 skb_pull_rcsum(skb, nh_ofs);
1093 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
1094 err = handle_fragments(net, key, info->zone.id, skb);
1100 err = ovs_ct_commit(net, key, info, skb);
1102 err = ovs_ct_lookup(net, key, info, skb);
1104 skb_push(skb, nh_ofs);
1105 skb_postpush_rcsum(skb, skb->data, nh_ofs);
1111 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
1112 const struct sw_flow_key *key, bool log)
1114 struct nf_conntrack_helper *helper;
1115 struct nf_conn_help *help;
1117 helper = nf_conntrack_helper_try_module_get(name, info->family,
1120 OVS_NLERR(log, "Unknown helper \"%s\"", name);
1124 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
1126 nf_conntrack_helper_put(helper);
1130 rcu_assign_pointer(help->helper, helper);
1131 info->helper = helper;
1135 #ifdef CONFIG_NF_NAT_NEEDED
1136 static int parse_nat(const struct nlattr *attr,
1137 struct ovs_conntrack_info *info, bool log)
1141 bool have_ip_max = false;
1142 bool have_proto_max = false;
1143 bool ip_vers = (info->family == NFPROTO_IPV6);
1145 nla_for_each_nested(a, attr, rem) {
1146 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
1147 [OVS_NAT_ATTR_SRC] = {0, 0},
1148 [OVS_NAT_ATTR_DST] = {0, 0},
1149 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
1150 sizeof(struct in6_addr)},
1151 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
1152 sizeof(struct in6_addr)},
1153 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
1154 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
1155 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
1156 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
1157 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
1159 int type = nla_type(a);
1161 if (type > OVS_NAT_ATTR_MAX) {
1163 "Unknown NAT attribute (type=%d, max=%d).\n",
1164 type, OVS_NAT_ATTR_MAX);
1168 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
1170 "NAT attribute type %d has unexpected length (%d != %d).\n",
1172 ovs_nat_attr_lens[type][ip_vers]);
1177 case OVS_NAT_ATTR_SRC:
1178 case OVS_NAT_ATTR_DST:
1181 "Only one type of NAT may be specified.\n"
1185 info->nat |= OVS_CT_NAT;
1186 info->nat |= ((type == OVS_NAT_ATTR_SRC)
1187 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
1190 case OVS_NAT_ATTR_IP_MIN:
1191 nla_memcpy(&info->range.min_addr, a,
1192 sizeof(info->range.min_addr));
1193 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1196 case OVS_NAT_ATTR_IP_MAX:
1198 nla_memcpy(&info->range.max_addr, a,
1199 sizeof(info->range.max_addr));
1200 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1203 case OVS_NAT_ATTR_PROTO_MIN:
1204 info->range.min_proto.all = htons(nla_get_u16(a));
1205 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1208 case OVS_NAT_ATTR_PROTO_MAX:
1209 have_proto_max = true;
1210 info->range.max_proto.all = htons(nla_get_u16(a));
1211 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1214 case OVS_NAT_ATTR_PERSISTENT:
1215 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1218 case OVS_NAT_ATTR_PROTO_HASH:
1219 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1222 case OVS_NAT_ATTR_PROTO_RANDOM:
1223 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1227 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1233 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1237 /* Do not allow flags if no type is given. */
1238 if (info->range.flags) {
1240 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1244 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1245 } else if (!info->commit) {
1247 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1251 /* Allow missing IP_MAX. */
1252 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1253 memcpy(&info->range.max_addr, &info->range.min_addr,
1254 sizeof(info->range.max_addr));
1256 /* Allow missing PROTO_MAX. */
1257 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1259 info->range.max_proto.all = info->range.min_proto.all;
1265 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1266 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
1267 [OVS_CT_ATTR_FORCE_COMMIT] = { .minlen = 0, .maxlen = 0 },
1268 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1269 .maxlen = sizeof(u16) },
1270 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1271 .maxlen = sizeof(struct md_mark) },
1272 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1273 .maxlen = sizeof(struct md_labels) },
1274 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
1275 .maxlen = NF_CT_HELPER_NAME_LEN },
1276 #ifdef CONFIG_NF_NAT_NEEDED
1277 /* NAT length is checked when parsing the nested attributes. */
1278 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1280 [OVS_CT_ATTR_EVENTMASK] = { .minlen = sizeof(u32),
1281 .maxlen = sizeof(u32) },
1284 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1285 const char **helper, bool log)
1290 nla_for_each_nested(a, attr, rem) {
1291 int type = nla_type(a);
1292 int maxlen = ovs_ct_attr_lens[type].maxlen;
1293 int minlen = ovs_ct_attr_lens[type].minlen;
1295 if (type > OVS_CT_ATTR_MAX) {
1297 "Unknown conntrack attr (type=%d, max=%d)",
1298 type, OVS_CT_ATTR_MAX);
1301 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1303 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1304 type, nla_len(a), maxlen);
1309 case OVS_CT_ATTR_FORCE_COMMIT:
1312 case OVS_CT_ATTR_COMMIT:
1313 info->commit = true;
1315 #ifdef CONFIG_NF_CONNTRACK_ZONES
1316 case OVS_CT_ATTR_ZONE:
1317 info->zone.id = nla_get_u16(a);
1320 #ifdef CONFIG_NF_CONNTRACK_MARK
1321 case OVS_CT_ATTR_MARK: {
1322 struct md_mark *mark = nla_data(a);
1325 OVS_NLERR(log, "ct_mark mask cannot be 0");
1332 #ifdef CONFIG_NF_CONNTRACK_LABELS
1333 case OVS_CT_ATTR_LABELS: {
1334 struct md_labels *labels = nla_data(a);
1336 if (!labels_nonzero(&labels->mask)) {
1337 OVS_NLERR(log, "ct_labels mask cannot be 0");
1340 info->labels = *labels;
1344 case OVS_CT_ATTR_HELPER:
1345 *helper = nla_data(a);
1346 if (!memchr(*helper, '\0', nla_len(a))) {
1347 OVS_NLERR(log, "Invalid conntrack helper");
1351 #ifdef CONFIG_NF_NAT_NEEDED
1352 case OVS_CT_ATTR_NAT: {
1353 int err = parse_nat(a, info, log);
1360 case OVS_CT_ATTR_EVENTMASK:
1361 info->have_eventmask = true;
1362 info->eventmask = nla_get_u32(a);
1366 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1372 #ifdef CONFIG_NF_CONNTRACK_MARK
1373 if (!info->commit && info->mark.mask) {
1375 "Setting conntrack mark requires 'commit' flag.");
1379 #ifdef CONFIG_NF_CONNTRACK_LABELS
1380 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1382 "Setting conntrack labels requires 'commit' flag.");
1387 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1394 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1396 if (attr == OVS_KEY_ATTR_CT_STATE)
1398 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1399 attr == OVS_KEY_ATTR_CT_ZONE)
1401 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1402 attr == OVS_KEY_ATTR_CT_MARK)
1404 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1405 attr == OVS_KEY_ATTR_CT_LABELS) {
1406 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1408 return ovs_net->xt_label;
1414 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1415 const struct sw_flow_key *key,
1416 struct sw_flow_actions **sfa, bool log)
1418 struct ovs_conntrack_info ct_info;
1419 const char *helper = NULL;
1423 family = key_to_nfproto(key);
1424 if (family == NFPROTO_UNSPEC) {
1425 OVS_NLERR(log, "ct family unspecified");
1429 memset(&ct_info, 0, sizeof(ct_info));
1430 ct_info.family = family;
1432 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1433 NF_CT_DEFAULT_ZONE_DIR, 0);
1435 err = parse_ct(attr, &ct_info, &helper, log);
1439 /* Set up template for tracking connections in specific zones. */
1440 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1442 OVS_NLERR(log, "Failed to allocate conntrack template");
1446 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1447 nf_conntrack_get(&ct_info.ct->ct_general);
1450 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1455 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1456 sizeof(ct_info), log);
1462 __ovs_ct_free_action(&ct_info);
1466 #ifdef CONFIG_NF_NAT_NEEDED
1467 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1468 struct sk_buff *skb)
1470 struct nlattr *start;
1472 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1476 if (info->nat & OVS_CT_SRC_NAT) {
1477 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1479 } else if (info->nat & OVS_CT_DST_NAT) {
1480 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1486 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1487 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1488 info->family == NFPROTO_IPV4) {
1489 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1490 info->range.min_addr.ip) ||
1491 (info->range.max_addr.ip
1492 != info->range.min_addr.ip &&
1493 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1494 info->range.max_addr.ip))))
1496 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1497 info->family == NFPROTO_IPV6) {
1498 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1499 &info->range.min_addr.in6) ||
1500 (memcmp(&info->range.max_addr.in6,
1501 &info->range.min_addr.in6,
1502 sizeof(info->range.max_addr.in6)) &&
1503 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1504 &info->range.max_addr.in6))))
1510 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1511 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1512 ntohs(info->range.min_proto.all)) ||
1513 (info->range.max_proto.all != info->range.min_proto.all &&
1514 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1515 ntohs(info->range.max_proto.all)))))
1518 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1519 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1521 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1522 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1524 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1525 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1528 nla_nest_end(skb, start);
1534 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1535 struct sk_buff *skb)
1537 struct nlattr *start;
1539 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1543 if (ct_info->commit && nla_put_flag(skb, ct_info->force
1544 ? OVS_CT_ATTR_FORCE_COMMIT
1545 : OVS_CT_ATTR_COMMIT))
1547 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1548 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1550 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1551 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1554 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1555 labels_nonzero(&ct_info->labels.mask) &&
1556 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1559 if (ct_info->helper) {
1560 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1561 ct_info->helper->name))
1564 if (ct_info->have_eventmask &&
1565 nla_put_u32(skb, OVS_CT_ATTR_EVENTMASK, ct_info->eventmask))
1568 #ifdef CONFIG_NF_NAT_NEEDED
1569 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1572 nla_nest_end(skb, start);
1577 void ovs_ct_free_action(const struct nlattr *a)
1579 struct ovs_conntrack_info *ct_info = nla_data(a);
1581 __ovs_ct_free_action(ct_info);
1584 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1586 if (ct_info->helper)
1587 nf_conntrack_helper_put(ct_info->helper);
1589 nf_ct_tmpl_free(ct_info->ct);
1592 void ovs_ct_init(struct net *net)
1594 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
1595 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1597 if (nf_connlabels_get(net, n_bits - 1)) {
1598 ovs_net->xt_label = false;
1599 OVS_NLERR(true, "Failed to set connlabel length");
1601 ovs_net->xt_label = true;
1605 void ovs_ct_exit(struct net *net)
1607 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1609 if (ovs_net->xt_label)
1610 nf_connlabels_put(net);