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 */
70 struct md_labels labels;
71 #ifdef CONFIG_NF_NAT_NEEDED
72 struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */
76 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
78 static u16 key_to_nfproto(const struct sw_flow_key *key)
80 switch (ntohs(key->eth.type)) {
86 return NFPROTO_UNSPEC;
90 /* Map SKB connection state into the values used by flow definition. */
91 static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
93 u8 ct_state = OVS_CS_F_TRACKED;
96 case IP_CT_ESTABLISHED_REPLY:
97 case IP_CT_RELATED_REPLY:
98 ct_state |= OVS_CS_F_REPLY_DIR;
105 case IP_CT_ESTABLISHED:
106 case IP_CT_ESTABLISHED_REPLY:
107 ct_state |= OVS_CS_F_ESTABLISHED;
110 case IP_CT_RELATED_REPLY:
111 ct_state |= OVS_CS_F_RELATED;
114 ct_state |= OVS_CS_F_NEW;
123 static u32 ovs_ct_get_mark(const struct nf_conn *ct)
125 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
126 return ct ? ct->mark : 0;
132 static void ovs_ct_get_labels(const struct nf_conn *ct,
133 struct ovs_key_ct_labels *labels)
135 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
138 size_t len = cl->words * sizeof(long);
140 if (len > OVS_CT_LABELS_LEN)
141 len = OVS_CT_LABELS_LEN;
142 else if (len < OVS_CT_LABELS_LEN)
143 memset(labels, 0, OVS_CT_LABELS_LEN);
144 memcpy(labels, cl->bits, len);
146 memset(labels, 0, OVS_CT_LABELS_LEN);
150 static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
151 const struct nf_conntrack_zone *zone,
152 const struct nf_conn *ct)
154 key->ct.state = state;
155 key->ct.zone = zone->id;
156 key->ct.mark = ovs_ct_get_mark(ct);
157 ovs_ct_get_labels(ct, &key->ct.labels);
160 /* Update 'key' based on skb->nfct. If 'post_ct' is true, then OVS has
161 * previously sent the packet to conntrack via the ct action. If
162 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
163 * initialized from the connection status.
165 static void ovs_ct_update_key(const struct sk_buff *skb,
166 const struct ovs_conntrack_info *info,
167 struct sw_flow_key *key, bool post_ct,
170 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
171 enum ip_conntrack_info ctinfo;
175 ct = nf_ct_get(skb, &ctinfo);
177 state = ovs_ct_get_state(ctinfo);
178 /* All unconfirmed entries are NEW connections. */
179 if (!nf_ct_is_confirmed(ct))
180 state |= OVS_CS_F_NEW;
181 /* OVS persists the related flag for the duration of the
185 state |= OVS_CS_F_RELATED;
186 if (keep_nat_flags) {
187 state |= key->ct.state & OVS_CS_F_NAT_MASK;
189 if (ct->status & IPS_SRC_NAT)
190 state |= OVS_CS_F_SRC_NAT;
191 if (ct->status & IPS_DST_NAT)
192 state |= OVS_CS_F_DST_NAT;
194 zone = nf_ct_zone(ct);
195 } else if (post_ct) {
196 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
200 __ovs_ct_update_key(key, state, zone, ct);
203 /* This is called to initialize CT key fields possibly coming in from the local
206 void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
208 ovs_ct_update_key(skb, NULL, key, false, false);
211 int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb)
213 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state))
216 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
217 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone))
220 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
221 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark))
224 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
225 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels),
232 static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key,
233 u32 ct_mark, u32 mask)
235 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
236 enum ip_conntrack_info ctinfo;
240 /* The connection could be invalid, in which case set_mark is no-op. */
241 ct = nf_ct_get(skb, &ctinfo);
245 new_mark = ct_mark | (ct->mark & ~(mask));
246 if (ct->mark != new_mark) {
248 nf_conntrack_event_cache(IPCT_MARK, ct);
249 key->ct.mark = new_mark;
258 static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key,
259 const struct ovs_key_ct_labels *labels,
260 const struct ovs_key_ct_labels *mask)
262 enum ip_conntrack_info ctinfo;
263 struct nf_conn_labels *cl;
267 /* The connection could be invalid, in which case set_label is no-op.*/
268 ct = nf_ct_get(skb, &ctinfo);
272 cl = nf_ct_labels_find(ct);
274 nf_ct_labels_ext_add(ct);
275 cl = nf_ct_labels_find(ct);
277 if (!cl || cl->words * sizeof(long) < OVS_CT_LABELS_LEN)
280 err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask,
281 OVS_CT_LABELS_LEN / sizeof(u32));
285 ovs_ct_get_labels(ct, &key->ct.labels);
289 /* 'skb' should already be pulled to nh_ofs. */
290 static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
292 const struct nf_conntrack_helper *helper;
293 const struct nf_conn_help *help;
294 enum ip_conntrack_info ctinfo;
295 unsigned int protoff;
299 ct = nf_ct_get(skb, &ctinfo);
300 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
303 help = nfct_help(ct);
307 helper = rcu_dereference(help->helper);
313 protoff = ip_hdrlen(skb);
316 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
320 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
322 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
323 pr_debug("proto header not found\n");
330 WARN_ONCE(1, "helper invoked on non-IP family!");
334 err = helper->help(skb, protoff, ct, ctinfo);
335 if (err != NF_ACCEPT)
338 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
339 * FTP with NAT) adusting the TCP payload size when mangling IP
340 * addresses and/or port numbers in the text-based control connection.
342 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
343 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
348 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
349 * value if 'skb' is freed.
351 static int handle_fragments(struct net *net, struct sw_flow_key *key,
352 u16 zone, struct sk_buff *skb)
354 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
357 if (key->eth.type == htons(ETH_P_IP)) {
358 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
360 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
361 err = ip_defrag(net, skb, user);
365 ovs_cb.mru = IPCB(skb)->frag_max_size;
366 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
367 } else if (key->eth.type == htons(ETH_P_IPV6)) {
368 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
371 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
372 err = nf_ct_frag6_gather(net, skb, user);
376 key->ip.proto = ipv6_hdr(skb)->nexthdr;
377 ovs_cb.mru = IP6CB(skb)->frag_max_size;
381 return -EPFNOSUPPORT;
384 key->ip.frag = OVS_FRAG_TYPE_NONE;
387 *OVS_CB(skb) = ovs_cb;
392 static struct nf_conntrack_expect *
393 ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
394 u16 proto, const struct sk_buff *skb)
396 struct nf_conntrack_tuple tuple;
398 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
400 return __nf_ct_expect_find(net, zone, &tuple);
403 /* This replicates logic from nf_conntrack_core.c that is not exported. */
404 static enum ip_conntrack_info
405 ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
407 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
409 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
410 return IP_CT_ESTABLISHED_REPLY;
411 /* Once we've had two way comms, always ESTABLISHED. */
412 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
413 return IP_CT_ESTABLISHED;
414 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
415 return IP_CT_RELATED;
419 /* Find an existing connection which this packet belongs to without
420 * re-attributing statistics or modifying the connection state. This allows an
421 * skb->nfct lost due to an upcall to be recovered during actions execution.
423 * Must be called with rcu_read_lock.
425 * On success, populates skb->nfct and skb->nfctinfo, and returns the
426 * connection. Returns NULL if there is no existing entry.
428 static struct nf_conn *
429 ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
430 u8 l3num, struct sk_buff *skb)
432 struct nf_conntrack_l3proto *l3proto;
433 struct nf_conntrack_l4proto *l4proto;
434 struct nf_conntrack_tuple tuple;
435 struct nf_conntrack_tuple_hash *h;
436 enum ip_conntrack_info ctinfo;
438 unsigned int dataoff;
441 l3proto = __nf_ct_l3proto_find(l3num);
443 pr_debug("ovs_ct_find_existing: Can't get l3proto\n");
446 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
448 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
451 l4proto = __nf_ct_l4proto_find(l3num, protonum);
453 pr_debug("ovs_ct_find_existing: Can't get l4proto\n");
456 if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
457 protonum, net, &tuple, l3proto, l4proto)) {
458 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
462 /* look for tuple match */
463 h = nf_conntrack_find_get(net, zone, &tuple);
465 return NULL; /* Not found. */
467 ct = nf_ct_tuplehash_to_ctrack(h);
469 ctinfo = ovs_ct_get_info(h);
470 if (ctinfo == IP_CT_NEW) {
471 /* This should not happen. */
472 WARN_ONCE(1, "ovs_ct_find_existing: new packet for %p\n", ct);
474 skb->nfct = &ct->ct_general;
475 skb->nfctinfo = ctinfo;
479 /* Determine whether skb->nfct is equal to the result of conntrack lookup. */
480 static bool skb_nfct_cached(struct net *net,
481 const struct sw_flow_key *key,
482 const struct ovs_conntrack_info *info,
485 enum ip_conntrack_info ctinfo;
488 ct = nf_ct_get(skb, &ctinfo);
489 /* If no ct, check if we have evidence that an existing conntrack entry
490 * might be found for this skb. This happens when we lose a skb->nfct
491 * due to an upcall. If the connection was not confirmed, it is not
492 * cached and needs to be run through conntrack again.
494 if (!ct && key->ct.state & OVS_CS_F_TRACKED &&
495 !(key->ct.state & OVS_CS_F_INVALID) &&
496 key->ct.zone == info->zone.id)
497 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb);
500 if (!net_eq(net, read_pnet(&ct->ct_net)))
502 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
505 struct nf_conn_help *help;
507 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
508 if (help && rcu_access_pointer(help->helper) != info->helper)
515 #ifdef CONFIG_NF_NAT_NEEDED
516 /* Modelled after nf_nat_ipv[46]_fn().
517 * range is only used for new, uninitialized NAT state.
518 * Returns either NF_ACCEPT or NF_DROP.
520 static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
521 enum ip_conntrack_info ctinfo,
522 const struct nf_nat_range *range,
523 enum nf_nat_manip_type maniptype)
525 int hooknum, nh_off, err = NF_ACCEPT;
527 nh_off = skb_network_offset(skb);
528 skb_pull(skb, nh_off);
530 /* See HOOK2MANIP(). */
531 if (maniptype == NF_NAT_MANIP_SRC)
532 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
534 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
538 case IP_CT_RELATED_REPLY:
539 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
540 skb->protocol == htons(ETH_P_IP) &&
541 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
542 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
546 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
547 skb->protocol == htons(ETH_P_IPV6)) {
549 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
550 int hdrlen = ipv6_skip_exthdr(skb,
551 sizeof(struct ipv6hdr),
552 &nexthdr, &frag_off);
554 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
555 if (!nf_nat_icmpv6_reply_translation(skb, ct,
563 /* Non-ICMP, fall thru to initialize if needed. */
565 /* Seen it before? This can happen for loopback, retrans,
568 if (!nf_nat_initialized(ct, maniptype)) {
569 /* Initialize according to the NAT action. */
570 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
571 /* Action is set up to establish a new
574 ? nf_nat_setup_info(ct, range, maniptype)
575 : nf_nat_alloc_null_binding(ct, hooknum);
576 if (err != NF_ACCEPT)
581 case IP_CT_ESTABLISHED:
582 case IP_CT_ESTABLISHED_REPLY:
590 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
592 skb_push(skb, nh_off);
597 static void ovs_nat_update_key(struct sw_flow_key *key,
598 const struct sk_buff *skb,
599 enum nf_nat_manip_type maniptype)
601 if (maniptype == NF_NAT_MANIP_SRC) {
604 key->ct.state |= OVS_CS_F_SRC_NAT;
605 if (key->eth.type == htons(ETH_P_IP))
606 key->ipv4.addr.src = ip_hdr(skb)->saddr;
607 else if (key->eth.type == htons(ETH_P_IPV6))
608 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
609 sizeof(key->ipv6.addr.src));
613 if (key->ip.proto == IPPROTO_UDP)
614 src = udp_hdr(skb)->source;
615 else if (key->ip.proto == IPPROTO_TCP)
616 src = tcp_hdr(skb)->source;
617 else if (key->ip.proto == IPPROTO_SCTP)
618 src = sctp_hdr(skb)->source;
626 key->ct.state |= OVS_CS_F_DST_NAT;
627 if (key->eth.type == htons(ETH_P_IP))
628 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
629 else if (key->eth.type == htons(ETH_P_IPV6))
630 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
631 sizeof(key->ipv6.addr.dst));
635 if (key->ip.proto == IPPROTO_UDP)
636 dst = udp_hdr(skb)->dest;
637 else if (key->ip.proto == IPPROTO_TCP)
638 dst = tcp_hdr(skb)->dest;
639 else if (key->ip.proto == IPPROTO_SCTP)
640 dst = sctp_hdr(skb)->dest;
648 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
649 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
650 const struct ovs_conntrack_info *info,
651 struct sk_buff *skb, struct nf_conn *ct,
652 enum ip_conntrack_info ctinfo)
654 enum nf_nat_manip_type maniptype;
657 if (nf_ct_is_untracked(ct)) {
658 /* A NAT action may only be performed on tracked packets. */
662 /* Add NAT extension if not confirmed yet. */
663 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
664 return NF_ACCEPT; /* Can't NAT. */
666 /* Determine NAT type.
667 * Check if the NAT type can be deduced from the tracked connection.
668 * Make sure new expected connections (IP_CT_RELATED) are NATted only
671 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
672 ct->status & IPS_NAT_MASK &&
673 (ctinfo != IP_CT_RELATED || info->commit)) {
674 /* NAT an established or related connection like before. */
675 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
676 /* This is the REPLY direction for a connection
677 * for which NAT was applied in the forward
678 * direction. Do the reverse NAT.
680 maniptype = ct->status & IPS_SRC_NAT
681 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
683 maniptype = ct->status & IPS_SRC_NAT
684 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
685 } else if (info->nat & OVS_CT_SRC_NAT) {
686 maniptype = NF_NAT_MANIP_SRC;
687 } else if (info->nat & OVS_CT_DST_NAT) {
688 maniptype = NF_NAT_MANIP_DST;
690 return NF_ACCEPT; /* Connection is not NATed. */
692 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
694 /* Mark NAT done if successful and update the flow key. */
695 if (err == NF_ACCEPT)
696 ovs_nat_update_key(key, skb, maniptype);
700 #else /* !CONFIG_NF_NAT_NEEDED */
701 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
702 const struct ovs_conntrack_info *info,
703 struct sk_buff *skb, struct nf_conn *ct,
704 enum ip_conntrack_info ctinfo)
710 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
711 * not done already. Update key with new CT state after passing the packet
713 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
714 * set to NULL and 0 will be returned.
716 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
717 const struct ovs_conntrack_info *info,
720 /* If we are recirculating packets to match on conntrack fields and
721 * committing with a separate conntrack action, then we don't need to
722 * actually run the packet through conntrack twice unless it's for a
725 bool cached = skb_nfct_cached(net, key, info, skb);
726 enum ip_conntrack_info ctinfo;
730 struct nf_conn *tmpl = info->ct;
733 /* Associate skb with specified zone. */
736 nf_conntrack_put(skb->nfct);
737 nf_conntrack_get(&tmpl->ct_general);
738 skb->nfct = &tmpl->ct_general;
739 skb->nfctinfo = IP_CT_NEW;
742 /* Repeat if requested, see nf_iterate(). */
744 err = nf_conntrack_in(net, info->family,
745 NF_INET_PRE_ROUTING, skb);
746 } while (err == NF_REPEAT);
748 if (err != NF_ACCEPT)
751 /* Clear CT state NAT flags to mark that we have not yet done
752 * NAT after the nf_conntrack_in() call. We can actually clear
753 * the whole state, as it will be re-initialized below.
757 /* Update the key, but keep the NAT flags. */
758 ovs_ct_update_key(skb, info, key, true, true);
761 ct = nf_ct_get(skb, &ctinfo);
763 /* Packets starting a new connection must be NATted before the
764 * helper, so that the helper knows about the NAT. We enforce
765 * this by delaying both NAT and helper calls for unconfirmed
766 * connections until the committing CT action. For later
767 * packets NAT and Helper may be called in either order.
769 * NAT will be done only if the CT action has NAT, and only
770 * once per packet (per zone), as guarded by the NAT bits in
773 if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) &&
774 (nf_ct_is_confirmed(ct) || info->commit) &&
775 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
779 /* Call the helper only if:
780 * - nf_conntrack_in() was executed above ("!cached") for a
781 * confirmed connection, or
782 * - When committing an unconfirmed connection.
784 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
785 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
793 /* Lookup connection and read fields into key. */
794 static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
795 const struct ovs_conntrack_info *info,
798 struct nf_conntrack_expect *exp;
800 /* If we pass an expected packet through nf_conntrack_in() the
801 * expectation is typically removed, but the packet could still be
802 * lost in upcall processing. To prevent this from happening we
803 * perform an explicit expectation lookup. Expected connections are
804 * always new, and will be passed through conntrack only when they are
805 * committed, as it is OK to remove the expectation at that time.
807 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
811 /* NOTE: New connections are NATted and Helped only when
812 * committed, so we are not calling into NAT here.
814 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
815 __ovs_ct_update_key(key, state, &info->zone, exp->master);
817 return __ovs_ct_lookup(net, key, info, skb);
822 /* Lookup connection and confirm if unconfirmed. */
823 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
824 const struct ovs_conntrack_info *info,
829 err = __ovs_ct_lookup(net, key, info, skb);
832 /* This is a no-op if the connection has already been confirmed. */
833 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
839 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
843 for (i = 0; i < sizeof(*labels); i++)
844 if (labels->ct_labels[i])
850 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
851 * value if 'skb' is freed.
853 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
854 struct sw_flow_key *key,
855 const struct ovs_conntrack_info *info)
860 /* The conntrack module expects to be working at L3. */
861 nh_ofs = skb_network_offset(skb);
862 skb_pull(skb, nh_ofs);
864 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
865 err = handle_fragments(net, key, info->zone.id, skb);
871 err = ovs_ct_commit(net, key, info, skb);
873 err = ovs_ct_lookup(net, key, info, skb);
877 if (info->mark.mask) {
878 err = ovs_ct_set_mark(skb, key, info->mark.value,
883 if (labels_nonzero(&info->labels.mask))
884 err = ovs_ct_set_labels(skb, key, &info->labels.value,
887 skb_push(skb, nh_ofs);
893 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
894 const struct sw_flow_key *key, bool log)
896 struct nf_conntrack_helper *helper;
897 struct nf_conn_help *help;
899 helper = nf_conntrack_helper_try_module_get(name, info->family,
902 OVS_NLERR(log, "Unknown helper \"%s\"", name);
906 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
908 module_put(helper->me);
912 rcu_assign_pointer(help->helper, helper);
913 info->helper = helper;
917 #ifdef CONFIG_NF_NAT_NEEDED
918 static int parse_nat(const struct nlattr *attr,
919 struct ovs_conntrack_info *info, bool log)
923 bool have_ip_max = false;
924 bool have_proto_max = false;
925 bool ip_vers = (info->family == NFPROTO_IPV6);
927 nla_for_each_nested(a, attr, rem) {
928 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
929 [OVS_NAT_ATTR_SRC] = {0, 0},
930 [OVS_NAT_ATTR_DST] = {0, 0},
931 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
932 sizeof(struct in6_addr)},
933 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
934 sizeof(struct in6_addr)},
935 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
936 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
937 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
938 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
939 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
941 int type = nla_type(a);
943 if (type > OVS_NAT_ATTR_MAX) {
945 "Unknown NAT attribute (type=%d, max=%d).\n",
946 type, OVS_NAT_ATTR_MAX);
950 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
952 "NAT attribute type %d has unexpected length (%d != %d).\n",
954 ovs_nat_attr_lens[type][ip_vers]);
959 case OVS_NAT_ATTR_SRC:
960 case OVS_NAT_ATTR_DST:
963 "Only one type of NAT may be specified.\n"
967 info->nat |= OVS_CT_NAT;
968 info->nat |= ((type == OVS_NAT_ATTR_SRC)
969 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
972 case OVS_NAT_ATTR_IP_MIN:
973 nla_memcpy(&info->range.min_addr, a,
974 sizeof(info->range.min_addr));
975 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
978 case OVS_NAT_ATTR_IP_MAX:
980 nla_memcpy(&info->range.max_addr, a,
981 sizeof(info->range.max_addr));
982 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
985 case OVS_NAT_ATTR_PROTO_MIN:
986 info->range.min_proto.all = htons(nla_get_u16(a));
987 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
990 case OVS_NAT_ATTR_PROTO_MAX:
991 have_proto_max = true;
992 info->range.max_proto.all = htons(nla_get_u16(a));
993 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
996 case OVS_NAT_ATTR_PERSISTENT:
997 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1000 case OVS_NAT_ATTR_PROTO_HASH:
1001 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1004 case OVS_NAT_ATTR_PROTO_RANDOM:
1005 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1009 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1015 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1019 /* Do not allow flags if no type is given. */
1020 if (info->range.flags) {
1022 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1026 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1027 } else if (!info->commit) {
1029 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1033 /* Allow missing IP_MAX. */
1034 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1035 memcpy(&info->range.max_addr, &info->range.min_addr,
1036 sizeof(info->range.max_addr));
1038 /* Allow missing PROTO_MAX. */
1039 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1041 info->range.max_proto.all = info->range.min_proto.all;
1047 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1048 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
1049 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1050 .maxlen = sizeof(u16) },
1051 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1052 .maxlen = sizeof(struct md_mark) },
1053 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1054 .maxlen = sizeof(struct md_labels) },
1055 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
1056 .maxlen = NF_CT_HELPER_NAME_LEN },
1057 #ifdef CONFIG_NF_NAT_NEEDED
1058 /* NAT length is checked when parsing the nested attributes. */
1059 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1063 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1064 const char **helper, bool log)
1069 nla_for_each_nested(a, attr, rem) {
1070 int type = nla_type(a);
1071 int maxlen = ovs_ct_attr_lens[type].maxlen;
1072 int minlen = ovs_ct_attr_lens[type].minlen;
1074 if (type > OVS_CT_ATTR_MAX) {
1076 "Unknown conntrack attr (type=%d, max=%d)",
1077 type, OVS_CT_ATTR_MAX);
1080 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1082 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1083 type, nla_len(a), maxlen);
1088 case OVS_CT_ATTR_COMMIT:
1089 info->commit = true;
1091 #ifdef CONFIG_NF_CONNTRACK_ZONES
1092 case OVS_CT_ATTR_ZONE:
1093 info->zone.id = nla_get_u16(a);
1096 #ifdef CONFIG_NF_CONNTRACK_MARK
1097 case OVS_CT_ATTR_MARK: {
1098 struct md_mark *mark = nla_data(a);
1101 OVS_NLERR(log, "ct_mark mask cannot be 0");
1108 #ifdef CONFIG_NF_CONNTRACK_LABELS
1109 case OVS_CT_ATTR_LABELS: {
1110 struct md_labels *labels = nla_data(a);
1112 if (!labels_nonzero(&labels->mask)) {
1113 OVS_NLERR(log, "ct_labels mask cannot be 0");
1116 info->labels = *labels;
1120 case OVS_CT_ATTR_HELPER:
1121 *helper = nla_data(a);
1122 if (!memchr(*helper, '\0', nla_len(a))) {
1123 OVS_NLERR(log, "Invalid conntrack helper");
1127 #ifdef CONFIG_NF_NAT_NEEDED
1128 case OVS_CT_ATTR_NAT: {
1129 int err = parse_nat(a, info, log);
1137 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1144 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1151 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1153 if (attr == OVS_KEY_ATTR_CT_STATE)
1155 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1156 attr == OVS_KEY_ATTR_CT_ZONE)
1158 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1159 attr == OVS_KEY_ATTR_CT_MARK)
1161 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1162 attr == OVS_KEY_ATTR_CT_LABELS) {
1163 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1165 return ovs_net->xt_label;
1171 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1172 const struct sw_flow_key *key,
1173 struct sw_flow_actions **sfa, bool log)
1175 struct ovs_conntrack_info ct_info;
1176 const char *helper = NULL;
1180 family = key_to_nfproto(key);
1181 if (family == NFPROTO_UNSPEC) {
1182 OVS_NLERR(log, "ct family unspecified");
1186 memset(&ct_info, 0, sizeof(ct_info));
1187 ct_info.family = family;
1189 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1190 NF_CT_DEFAULT_ZONE_DIR, 0);
1192 err = parse_ct(attr, &ct_info, &helper, log);
1196 /* Set up template for tracking connections in specific zones. */
1197 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1199 OVS_NLERR(log, "Failed to allocate conntrack template");
1203 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1204 nf_conntrack_get(&ct_info.ct->ct_general);
1207 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1212 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1213 sizeof(ct_info), log);
1219 __ovs_ct_free_action(&ct_info);
1223 #ifdef CONFIG_NF_NAT_NEEDED
1224 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1225 struct sk_buff *skb)
1227 struct nlattr *start;
1229 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1233 if (info->nat & OVS_CT_SRC_NAT) {
1234 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1236 } else if (info->nat & OVS_CT_DST_NAT) {
1237 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1243 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1244 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1245 info->family == NFPROTO_IPV4) {
1246 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1247 info->range.min_addr.ip) ||
1248 (info->range.max_addr.ip
1249 != info->range.min_addr.ip &&
1250 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1251 info->range.max_addr.ip))))
1253 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1254 info->family == NFPROTO_IPV6) {
1255 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1256 &info->range.min_addr.in6) ||
1257 (memcmp(&info->range.max_addr.in6,
1258 &info->range.min_addr.in6,
1259 sizeof(info->range.max_addr.in6)) &&
1260 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1261 &info->range.max_addr.in6))))
1267 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1268 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1269 ntohs(info->range.min_proto.all)) ||
1270 (info->range.max_proto.all != info->range.min_proto.all &&
1271 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1272 ntohs(info->range.max_proto.all)))))
1275 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1276 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1278 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1279 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1281 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1282 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1285 nla_nest_end(skb, start);
1291 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1292 struct sk_buff *skb)
1294 struct nlattr *start;
1296 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1300 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
1302 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1303 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1305 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1306 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1309 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1310 labels_nonzero(&ct_info->labels.mask) &&
1311 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1314 if (ct_info->helper) {
1315 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1316 ct_info->helper->name))
1319 #ifdef CONFIG_NF_NAT_NEEDED
1320 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1323 nla_nest_end(skb, start);
1328 void ovs_ct_free_action(const struct nlattr *a)
1330 struct ovs_conntrack_info *ct_info = nla_data(a);
1332 __ovs_ct_free_action(ct_info);
1335 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1337 if (ct_info->helper)
1338 module_put(ct_info->helper->me);
1340 nf_ct_put(ct_info->ct);
1343 void ovs_ct_init(struct net *net)
1345 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
1346 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1348 if (nf_connlabels_get(net, n_bits - 1)) {
1349 ovs_net->xt_label = false;
1350 OVS_NLERR(true, "Failed to set connlabel length");
1352 ovs_net->xt_label = true;
1356 void ovs_ct_exit(struct net *net)
1358 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1360 if (ovs_net->xt_label)
1361 nf_connlabels_put(net);