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>
19 #include <linux/static_key.h>
21 #include <net/genetlink.h>
22 #include <net/netfilter/nf_conntrack_core.h>
23 #include <net/netfilter/nf_conntrack_count.h>
24 #include <net/netfilter/nf_conntrack_helper.h>
25 #include <net/netfilter/nf_conntrack_labels.h>
26 #include <net/netfilter/nf_conntrack_seqadj.h>
27 #include <net/netfilter/nf_conntrack_timeout.h>
28 #include <net/netfilter/nf_conntrack_zones.h>
29 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
30 #include <net/ipv6_frag.h>
32 #if IS_ENABLED(CONFIG_NF_NAT)
33 #include <net/netfilter/nf_nat.h>
37 #include "conntrack.h"
39 #include "flow_netlink.h"
41 struct ovs_ct_len_tbl {
46 /* Metadata mark for masked write to conntrack mark */
52 /* Metadata label for masked write to conntrack label. */
54 struct ovs_key_ct_labels value;
55 struct ovs_key_ct_labels mask;
59 OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */
60 OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */
61 OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */
64 /* Conntrack action context for execution. */
65 struct ovs_conntrack_info {
66 struct nf_conntrack_helper *helper;
67 struct nf_conntrack_zone zone;
70 u8 nat : 3; /* enum ovs_ct_nat */
72 u8 have_eventmask : 1;
74 u32 eventmask; /* Mask of 1 << IPCT_*. */
76 struct md_labels labels;
77 char timeout[CTNL_TIMEOUT_NAME_MAX];
78 #if IS_ENABLED(CONFIG_NF_NAT)
79 struct nf_nat_range2 range; /* Only present for SRC NAT and DST NAT. */
83 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
84 #define OVS_CT_LIMIT_UNLIMITED 0
85 #define OVS_CT_LIMIT_DEFAULT OVS_CT_LIMIT_UNLIMITED
86 #define CT_LIMIT_HASH_BUCKETS 512
87 static DEFINE_STATIC_KEY_FALSE(ovs_ct_limit_enabled);
90 /* Elements in ovs_ct_limit_info->limits hash table */
91 struct hlist_node hlist_node;
97 struct ovs_ct_limit_info {
99 struct hlist_head *limits;
100 struct nf_conncount_data *data;
103 static const struct nla_policy ct_limit_policy[OVS_CT_LIMIT_ATTR_MAX + 1] = {
104 [OVS_CT_LIMIT_ATTR_ZONE_LIMIT] = { .type = NLA_NESTED, },
108 static bool labels_nonzero(const struct ovs_key_ct_labels *labels);
110 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
112 static u16 key_to_nfproto(const struct sw_flow_key *key)
114 switch (ntohs(key->eth.type)) {
120 return NFPROTO_UNSPEC;
124 /* Map SKB connection state into the values used by flow definition. */
125 static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
127 u8 ct_state = OVS_CS_F_TRACKED;
130 case IP_CT_ESTABLISHED_REPLY:
131 case IP_CT_RELATED_REPLY:
132 ct_state |= OVS_CS_F_REPLY_DIR;
139 case IP_CT_ESTABLISHED:
140 case IP_CT_ESTABLISHED_REPLY:
141 ct_state |= OVS_CS_F_ESTABLISHED;
144 case IP_CT_RELATED_REPLY:
145 ct_state |= OVS_CS_F_RELATED;
148 ct_state |= OVS_CS_F_NEW;
157 static u32 ovs_ct_get_mark(const struct nf_conn *ct)
159 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
160 return ct ? ct->mark : 0;
166 /* Guard against conntrack labels max size shrinking below 128 bits. */
167 #if NF_CT_LABELS_MAX_SIZE < 16
168 #error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes
171 static void ovs_ct_get_labels(const struct nf_conn *ct,
172 struct ovs_key_ct_labels *labels)
174 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
177 memcpy(labels, cl->bits, OVS_CT_LABELS_LEN);
179 memset(labels, 0, OVS_CT_LABELS_LEN);
182 static void __ovs_ct_update_key_orig_tp(struct sw_flow_key *key,
183 const struct nf_conntrack_tuple *orig,
186 key->ct_orig_proto = orig->dst.protonum;
187 if (orig->dst.protonum == icmp_proto) {
188 key->ct.orig_tp.src = htons(orig->dst.u.icmp.type);
189 key->ct.orig_tp.dst = htons(orig->dst.u.icmp.code);
191 key->ct.orig_tp.src = orig->src.u.all;
192 key->ct.orig_tp.dst = orig->dst.u.all;
196 static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
197 const struct nf_conntrack_zone *zone,
198 const struct nf_conn *ct)
200 key->ct_state = state;
201 key->ct_zone = zone->id;
202 key->ct.mark = ovs_ct_get_mark(ct);
203 ovs_ct_get_labels(ct, &key->ct.labels);
206 const struct nf_conntrack_tuple *orig;
208 /* Use the master if we have one. */
211 orig = &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple;
213 /* IP version must match with the master connection. */
214 if (key->eth.type == htons(ETH_P_IP) &&
215 nf_ct_l3num(ct) == NFPROTO_IPV4) {
216 key->ipv4.ct_orig.src = orig->src.u3.ip;
217 key->ipv4.ct_orig.dst = orig->dst.u3.ip;
218 __ovs_ct_update_key_orig_tp(key, orig, IPPROTO_ICMP);
220 } else if (key->eth.type == htons(ETH_P_IPV6) &&
221 !sw_flow_key_is_nd(key) &&
222 nf_ct_l3num(ct) == NFPROTO_IPV6) {
223 key->ipv6.ct_orig.src = orig->src.u3.in6;
224 key->ipv6.ct_orig.dst = orig->dst.u3.in6;
225 __ovs_ct_update_key_orig_tp(key, orig, NEXTHDR_ICMP);
229 /* Clear 'ct_orig_proto' to mark the non-existence of conntrack
230 * original direction key fields.
232 key->ct_orig_proto = 0;
235 /* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has
236 * previously sent the packet to conntrack via the ct action. If
237 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
238 * initialized from the connection status.
240 static void ovs_ct_update_key(const struct sk_buff *skb,
241 const struct ovs_conntrack_info *info,
242 struct sw_flow_key *key, bool post_ct,
245 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
246 enum ip_conntrack_info ctinfo;
250 ct = nf_ct_get(skb, &ctinfo);
252 state = ovs_ct_get_state(ctinfo);
253 /* All unconfirmed entries are NEW connections. */
254 if (!nf_ct_is_confirmed(ct))
255 state |= OVS_CS_F_NEW;
256 /* OVS persists the related flag for the duration of the
260 state |= OVS_CS_F_RELATED;
261 if (keep_nat_flags) {
262 state |= key->ct_state & OVS_CS_F_NAT_MASK;
264 if (ct->status & IPS_SRC_NAT)
265 state |= OVS_CS_F_SRC_NAT;
266 if (ct->status & IPS_DST_NAT)
267 state |= OVS_CS_F_DST_NAT;
269 zone = nf_ct_zone(ct);
270 } else if (post_ct) {
271 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
275 __ovs_ct_update_key(key, state, zone, ct);
278 /* This is called to initialize CT key fields possibly coming in from the local
281 void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
283 ovs_ct_update_key(skb, NULL, key, false, false);
286 #define IN6_ADDR_INITIALIZER(ADDR) \
287 { (ADDR).s6_addr32[0], (ADDR).s6_addr32[1], \
288 (ADDR).s6_addr32[2], (ADDR).s6_addr32[3] }
290 int ovs_ct_put_key(const struct sw_flow_key *swkey,
291 const struct sw_flow_key *output, struct sk_buff *skb)
293 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, output->ct_state))
296 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
297 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, output->ct_zone))
300 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
301 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, output->ct.mark))
304 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
305 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(output->ct.labels),
309 if (swkey->ct_orig_proto) {
310 if (swkey->eth.type == htons(ETH_P_IP)) {
311 struct ovs_key_ct_tuple_ipv4 orig = {
312 output->ipv4.ct_orig.src,
313 output->ipv4.ct_orig.dst,
314 output->ct.orig_tp.src,
315 output->ct.orig_tp.dst,
316 output->ct_orig_proto,
318 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4,
319 sizeof(orig), &orig))
321 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
322 struct ovs_key_ct_tuple_ipv6 orig = {
323 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.src),
324 IN6_ADDR_INITIALIZER(output->ipv6.ct_orig.dst),
325 output->ct.orig_tp.src,
326 output->ct.orig_tp.dst,
327 output->ct_orig_proto,
329 if (nla_put(skb, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6,
330 sizeof(orig), &orig))
338 static int ovs_ct_set_mark(struct nf_conn *ct, struct sw_flow_key *key,
339 u32 ct_mark, u32 mask)
341 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
344 new_mark = ct_mark | (ct->mark & ~(mask));
345 if (ct->mark != new_mark) {
347 if (nf_ct_is_confirmed(ct))
348 nf_conntrack_event_cache(IPCT_MARK, ct);
349 key->ct.mark = new_mark;
358 static struct nf_conn_labels *ovs_ct_get_conn_labels(struct nf_conn *ct)
360 struct nf_conn_labels *cl;
362 cl = nf_ct_labels_find(ct);
364 nf_ct_labels_ext_add(ct);
365 cl = nf_ct_labels_find(ct);
371 /* Initialize labels for a new, yet to be committed conntrack entry. Note that
372 * since the new connection is not yet confirmed, and thus no-one else has
373 * access to it's labels, we simply write them over.
375 static int ovs_ct_init_labels(struct nf_conn *ct, struct sw_flow_key *key,
376 const struct ovs_key_ct_labels *labels,
377 const struct ovs_key_ct_labels *mask)
379 struct nf_conn_labels *cl, *master_cl;
380 bool have_mask = labels_nonzero(mask);
382 /* Inherit master's labels to the related connection? */
383 master_cl = ct->master ? nf_ct_labels_find(ct->master) : NULL;
385 if (!master_cl && !have_mask)
386 return 0; /* Nothing to do. */
388 cl = ovs_ct_get_conn_labels(ct);
392 /* Inherit the master's labels, if any. */
397 u32 *dst = (u32 *)cl->bits;
400 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
401 dst[i] = (dst[i] & ~mask->ct_labels_32[i]) |
402 (labels->ct_labels_32[i]
403 & mask->ct_labels_32[i]);
406 /* Labels are included in the IPCTNL_MSG_CT_NEW event only if the
407 * IPCT_LABEL bit is set in the event cache.
409 nf_conntrack_event_cache(IPCT_LABEL, ct);
411 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
416 static int ovs_ct_set_labels(struct nf_conn *ct, struct sw_flow_key *key,
417 const struct ovs_key_ct_labels *labels,
418 const struct ovs_key_ct_labels *mask)
420 struct nf_conn_labels *cl;
423 cl = ovs_ct_get_conn_labels(ct);
427 err = nf_connlabels_replace(ct, labels->ct_labels_32,
429 OVS_CT_LABELS_LEN_32);
433 memcpy(&key->ct.labels, cl->bits, OVS_CT_LABELS_LEN);
438 /* 'skb' should already be pulled to nh_ofs. */
439 static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
441 const struct nf_conntrack_helper *helper;
442 const struct nf_conn_help *help;
443 enum ip_conntrack_info ctinfo;
444 unsigned int protoff;
448 ct = nf_ct_get(skb, &ctinfo);
449 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
452 help = nfct_help(ct);
456 helper = rcu_dereference(help->helper);
462 protoff = ip_hdrlen(skb);
465 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
469 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
471 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
472 pr_debug("proto header not found\n");
479 WARN_ONCE(1, "helper invoked on non-IP family!");
483 err = helper->help(skb, protoff, ct, ctinfo);
484 if (err != NF_ACCEPT)
487 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
488 * FTP with NAT) adusting the TCP payload size when mangling IP
489 * addresses and/or port numbers in the text-based control connection.
491 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
492 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
497 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
498 * value if 'skb' is freed.
500 static int handle_fragments(struct net *net, struct sw_flow_key *key,
501 u16 zone, struct sk_buff *skb)
503 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
506 if (key->eth.type == htons(ETH_P_IP)) {
507 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
509 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
510 err = ip_defrag(net, skb, user);
514 ovs_cb.mru = IPCB(skb)->frag_max_size;
515 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
516 } else if (key->eth.type == htons(ETH_P_IPV6)) {
517 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
519 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
520 err = nf_ct_frag6_gather(net, skb, user);
522 if (err != -EINPROGRESS)
527 key->ip.proto = ipv6_hdr(skb)->nexthdr;
528 ovs_cb.mru = IP6CB(skb)->frag_max_size;
532 return -EPFNOSUPPORT;
535 key->ip.frag = OVS_FRAG_TYPE_NONE;
538 *OVS_CB(skb) = ovs_cb;
543 static struct nf_conntrack_expect *
544 ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
545 u16 proto, const struct sk_buff *skb)
547 struct nf_conntrack_tuple tuple;
548 struct nf_conntrack_expect *exp;
550 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
553 exp = __nf_ct_expect_find(net, zone, &tuple);
555 struct nf_conntrack_tuple_hash *h;
557 /* Delete existing conntrack entry, if it clashes with the
558 * expectation. This can happen since conntrack ALGs do not
559 * check for clashes between (new) expectations and existing
560 * conntrack entries. nf_conntrack_in() will check the
561 * expectations only if a conntrack entry can not be found,
562 * which can lead to OVS finding the expectation (here) in the
563 * init direction, but which will not be removed by the
564 * nf_conntrack_in() call, if a matching conntrack entry is
565 * found instead. In this case all init direction packets
566 * would be reported as new related packets, while reply
567 * direction packets would be reported as un-related
568 * established packets.
570 h = nf_conntrack_find_get(net, zone, &tuple);
572 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
574 nf_ct_delete(ct, 0, 0);
575 nf_conntrack_put(&ct->ct_general);
582 /* This replicates logic from nf_conntrack_core.c that is not exported. */
583 static enum ip_conntrack_info
584 ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
586 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
588 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
589 return IP_CT_ESTABLISHED_REPLY;
590 /* Once we've had two way comms, always ESTABLISHED. */
591 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
592 return IP_CT_ESTABLISHED;
593 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
594 return IP_CT_RELATED;
598 /* Find an existing connection which this packet belongs to without
599 * re-attributing statistics or modifying the connection state. This allows an
600 * skb->_nfct lost due to an upcall to be recovered during actions execution.
602 * Must be called with rcu_read_lock.
604 * On success, populates skb->_nfct and returns the connection. Returns NULL
605 * if there is no existing entry.
607 static struct nf_conn *
608 ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
609 u8 l3num, struct sk_buff *skb, bool natted)
611 struct nf_conntrack_tuple tuple;
612 struct nf_conntrack_tuple_hash *h;
615 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), l3num,
617 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
621 /* Must invert the tuple if skb has been transformed by NAT. */
623 struct nf_conntrack_tuple inverse;
625 if (!nf_ct_invert_tuple(&inverse, &tuple)) {
626 pr_debug("ovs_ct_find_existing: Inversion failed!\n");
632 /* look for tuple match */
633 h = nf_conntrack_find_get(net, zone, &tuple);
635 return NULL; /* Not found. */
637 ct = nf_ct_tuplehash_to_ctrack(h);
639 /* Inverted packet tuple matches the reverse direction conntrack tuple,
640 * select the other tuplehash to get the right 'ctinfo' bits for this
644 h = &ct->tuplehash[!h->tuple.dst.dir];
646 nf_ct_set(skb, ct, ovs_ct_get_info(h));
651 struct nf_conn *ovs_ct_executed(struct net *net,
652 const struct sw_flow_key *key,
653 const struct ovs_conntrack_info *info,
657 struct nf_conn *ct = NULL;
659 /* If no ct, check if we have evidence that an existing conntrack entry
660 * might be found for this skb. This happens when we lose a skb->_nfct
661 * due to an upcall, or if the direction is being forced. If the
662 * connection was not confirmed, it is not cached and needs to be run
663 * through conntrack again.
665 *ct_executed = (key->ct_state & OVS_CS_F_TRACKED) &&
666 !(key->ct_state & OVS_CS_F_INVALID) &&
667 (key->ct_zone == info->zone.id);
669 if (*ct_executed || (!key->ct_state && info->force)) {
670 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb,
678 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
679 static bool skb_nfct_cached(struct net *net,
680 const struct sw_flow_key *key,
681 const struct ovs_conntrack_info *info,
684 enum ip_conntrack_info ctinfo;
686 bool ct_executed = true;
688 ct = nf_ct_get(skb, &ctinfo);
690 ct = ovs_ct_executed(net, key, info, skb, &ct_executed);
693 nf_ct_get(skb, &ctinfo);
697 if (!net_eq(net, read_pnet(&ct->ct_net)))
699 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
702 struct nf_conn_help *help;
704 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
705 if (help && rcu_access_pointer(help->helper) != info->helper)
708 /* Force conntrack entry direction to the current packet? */
709 if (info->force && CTINFO2DIR(ctinfo) != IP_CT_DIR_ORIGINAL) {
710 /* Delete the conntrack entry if confirmed, else just release
713 if (nf_ct_is_confirmed(ct))
714 nf_ct_delete(ct, 0, 0);
716 nf_conntrack_put(&ct->ct_general);
717 nf_ct_set(skb, NULL, 0);
724 #if IS_ENABLED(CONFIG_NF_NAT)
725 /* Modelled after nf_nat_ipv[46]_fn().
726 * range is only used for new, uninitialized NAT state.
727 * Returns either NF_ACCEPT or NF_DROP.
729 static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
730 enum ip_conntrack_info ctinfo,
731 const struct nf_nat_range2 *range,
732 enum nf_nat_manip_type maniptype)
734 int hooknum, nh_off, err = NF_ACCEPT;
736 nh_off = skb_network_offset(skb);
737 skb_pull_rcsum(skb, nh_off);
739 /* See HOOK2MANIP(). */
740 if (maniptype == NF_NAT_MANIP_SRC)
741 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
743 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
747 case IP_CT_RELATED_REPLY:
748 if (IS_ENABLED(CONFIG_NF_NAT) &&
749 skb->protocol == htons(ETH_P_IP) &&
750 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
751 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
755 } else if (IS_ENABLED(CONFIG_IPV6) &&
756 skb->protocol == htons(ETH_P_IPV6)) {
758 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
759 int hdrlen = ipv6_skip_exthdr(skb,
760 sizeof(struct ipv6hdr),
761 &nexthdr, &frag_off);
763 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
764 if (!nf_nat_icmpv6_reply_translation(skb, ct,
772 /* Non-ICMP, fall thru to initialize if needed. */
775 /* Seen it before? This can happen for loopback, retrans,
778 if (!nf_nat_initialized(ct, maniptype)) {
779 /* Initialize according to the NAT action. */
780 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
781 /* Action is set up to establish a new
784 ? nf_nat_setup_info(ct, range, maniptype)
785 : nf_nat_alloc_null_binding(ct, hooknum);
786 if (err != NF_ACCEPT)
791 case IP_CT_ESTABLISHED:
792 case IP_CT_ESTABLISHED_REPLY:
800 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
802 skb_push(skb, nh_off);
803 skb_postpush_rcsum(skb, skb->data, nh_off);
808 static void ovs_nat_update_key(struct sw_flow_key *key,
809 const struct sk_buff *skb,
810 enum nf_nat_manip_type maniptype)
812 if (maniptype == NF_NAT_MANIP_SRC) {
815 key->ct_state |= OVS_CS_F_SRC_NAT;
816 if (key->eth.type == htons(ETH_P_IP))
817 key->ipv4.addr.src = ip_hdr(skb)->saddr;
818 else if (key->eth.type == htons(ETH_P_IPV6))
819 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
820 sizeof(key->ipv6.addr.src));
824 if (key->ip.proto == IPPROTO_UDP)
825 src = udp_hdr(skb)->source;
826 else if (key->ip.proto == IPPROTO_TCP)
827 src = tcp_hdr(skb)->source;
828 else if (key->ip.proto == IPPROTO_SCTP)
829 src = sctp_hdr(skb)->source;
837 key->ct_state |= OVS_CS_F_DST_NAT;
838 if (key->eth.type == htons(ETH_P_IP))
839 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
840 else if (key->eth.type == htons(ETH_P_IPV6))
841 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
842 sizeof(key->ipv6.addr.dst));
846 if (key->ip.proto == IPPROTO_UDP)
847 dst = udp_hdr(skb)->dest;
848 else if (key->ip.proto == IPPROTO_TCP)
849 dst = tcp_hdr(skb)->dest;
850 else if (key->ip.proto == IPPROTO_SCTP)
851 dst = sctp_hdr(skb)->dest;
859 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
860 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
861 const struct ovs_conntrack_info *info,
862 struct sk_buff *skb, struct nf_conn *ct,
863 enum ip_conntrack_info ctinfo)
865 enum nf_nat_manip_type maniptype;
868 /* Add NAT extension if not confirmed yet. */
869 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
870 return NF_ACCEPT; /* Can't NAT. */
872 /* Determine NAT type.
873 * Check if the NAT type can be deduced from the tracked connection.
874 * Make sure new expected connections (IP_CT_RELATED) are NATted only
877 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
878 ct->status & IPS_NAT_MASK &&
879 (ctinfo != IP_CT_RELATED || info->commit)) {
880 /* NAT an established or related connection like before. */
881 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
882 /* This is the REPLY direction for a connection
883 * for which NAT was applied in the forward
884 * direction. Do the reverse NAT.
886 maniptype = ct->status & IPS_SRC_NAT
887 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
889 maniptype = ct->status & IPS_SRC_NAT
890 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
891 } else if (info->nat & OVS_CT_SRC_NAT) {
892 maniptype = NF_NAT_MANIP_SRC;
893 } else if (info->nat & OVS_CT_DST_NAT) {
894 maniptype = NF_NAT_MANIP_DST;
896 return NF_ACCEPT; /* Connection is not NATed. */
898 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
900 /* Mark NAT done if successful and update the flow key. */
901 if (err == NF_ACCEPT)
902 ovs_nat_update_key(key, skb, maniptype);
906 #else /* !CONFIG_NF_NAT */
907 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
908 const struct ovs_conntrack_info *info,
909 struct sk_buff *skb, struct nf_conn *ct,
910 enum ip_conntrack_info ctinfo)
916 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
917 * not done already. Update key with new CT state after passing the packet
919 * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be
920 * set to NULL and 0 will be returned.
922 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
923 const struct ovs_conntrack_info *info,
926 /* If we are recirculating packets to match on conntrack fields and
927 * committing with a separate conntrack action, then we don't need to
928 * actually run the packet through conntrack twice unless it's for a
931 bool cached = skb_nfct_cached(net, key, info, skb);
932 enum ip_conntrack_info ctinfo;
936 struct nf_hook_state state = {
937 .hook = NF_INET_PRE_ROUTING,
941 struct nf_conn *tmpl = info->ct;
944 /* Associate skb with specified zone. */
947 nf_conntrack_put(skb_nfct(skb));
948 nf_conntrack_get(&tmpl->ct_general);
949 nf_ct_set(skb, tmpl, IP_CT_NEW);
952 err = nf_conntrack_in(skb, &state);
953 if (err != NF_ACCEPT)
956 /* Clear CT state NAT flags to mark that we have not yet done
957 * NAT after the nf_conntrack_in() call. We can actually clear
958 * the whole state, as it will be re-initialized below.
962 /* Update the key, but keep the NAT flags. */
963 ovs_ct_update_key(skb, info, key, true, true);
966 ct = nf_ct_get(skb, &ctinfo);
968 /* Packets starting a new connection must be NATted before the
969 * helper, so that the helper knows about the NAT. We enforce
970 * this by delaying both NAT and helper calls for unconfirmed
971 * connections until the committing CT action. For later
972 * packets NAT and Helper may be called in either order.
974 * NAT will be done only if the CT action has NAT, and only
975 * once per packet (per zone), as guarded by the NAT bits in
978 if (info->nat && !(key->ct_state & OVS_CS_F_NAT_MASK) &&
979 (nf_ct_is_confirmed(ct) || info->commit) &&
980 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
984 /* Userspace may decide to perform a ct lookup without a helper
985 * specified followed by a (recirculate and) commit with one.
986 * Therefore, for unconfirmed connections which we will commit,
987 * we need to attach the helper here.
989 if (!nf_ct_is_confirmed(ct) && info->commit &&
990 info->helper && !nfct_help(ct)) {
991 int err = __nf_ct_try_assign_helper(ct, info->ct,
996 /* helper installed, add seqadj if NAT is required */
997 if (info->nat && !nfct_seqadj(ct)) {
998 if (!nfct_seqadj_ext_add(ct))
1003 /* Call the helper only if:
1004 * - nf_conntrack_in() was executed above ("!cached") for a
1005 * confirmed connection, or
1006 * - When committing an unconfirmed connection.
1008 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
1009 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
1017 /* Lookup connection and read fields into key. */
1018 static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
1019 const struct ovs_conntrack_info *info,
1020 struct sk_buff *skb)
1022 struct nf_conntrack_expect *exp;
1024 /* If we pass an expected packet through nf_conntrack_in() the
1025 * expectation is typically removed, but the packet could still be
1026 * lost in upcall processing. To prevent this from happening we
1027 * perform an explicit expectation lookup. Expected connections are
1028 * always new, and will be passed through conntrack only when they are
1029 * committed, as it is OK to remove the expectation at that time.
1031 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
1035 /* NOTE: New connections are NATted and Helped only when
1036 * committed, so we are not calling into NAT here.
1038 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
1039 __ovs_ct_update_key(key, state, &info->zone, exp->master);
1044 err = __ovs_ct_lookup(net, key, info, skb);
1048 ct = (struct nf_conn *)skb_nfct(skb);
1050 nf_ct_deliver_cached_events(ct);
1056 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
1060 for (i = 0; i < OVS_CT_LABELS_LEN_32; i++)
1061 if (labels->ct_labels_32[i])
1067 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1068 static struct hlist_head *ct_limit_hash_bucket(
1069 const struct ovs_ct_limit_info *info, u16 zone)
1071 return &info->limits[zone & (CT_LIMIT_HASH_BUCKETS - 1)];
1074 /* Call with ovs_mutex */
1075 static void ct_limit_set(const struct ovs_ct_limit_info *info,
1076 struct ovs_ct_limit *new_ct_limit)
1078 struct ovs_ct_limit *ct_limit;
1079 struct hlist_head *head;
1081 head = ct_limit_hash_bucket(info, new_ct_limit->zone);
1082 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
1083 if (ct_limit->zone == new_ct_limit->zone) {
1084 hlist_replace_rcu(&ct_limit->hlist_node,
1085 &new_ct_limit->hlist_node);
1086 kfree_rcu(ct_limit, rcu);
1091 hlist_add_head_rcu(&new_ct_limit->hlist_node, head);
1094 /* Call with ovs_mutex */
1095 static void ct_limit_del(const struct ovs_ct_limit_info *info, u16 zone)
1097 struct ovs_ct_limit *ct_limit;
1098 struct hlist_head *head;
1099 struct hlist_node *n;
1101 head = ct_limit_hash_bucket(info, zone);
1102 hlist_for_each_entry_safe(ct_limit, n, head, hlist_node) {
1103 if (ct_limit->zone == zone) {
1104 hlist_del_rcu(&ct_limit->hlist_node);
1105 kfree_rcu(ct_limit, rcu);
1111 /* Call with RCU read lock */
1112 static u32 ct_limit_get(const struct ovs_ct_limit_info *info, u16 zone)
1114 struct ovs_ct_limit *ct_limit;
1115 struct hlist_head *head;
1117 head = ct_limit_hash_bucket(info, zone);
1118 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
1119 if (ct_limit->zone == zone)
1120 return ct_limit->limit;
1123 return info->default_limit;
1126 static int ovs_ct_check_limit(struct net *net,
1127 const struct ovs_conntrack_info *info,
1128 const struct nf_conntrack_tuple *tuple)
1130 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1131 const struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
1132 u32 per_zone_limit, connections;
1135 conncount_key = info->zone.id;
1137 per_zone_limit = ct_limit_get(ct_limit_info, info->zone.id);
1138 if (per_zone_limit == OVS_CT_LIMIT_UNLIMITED)
1141 connections = nf_conncount_count(net, ct_limit_info->data,
1142 &conncount_key, tuple, &info->zone);
1143 if (connections > per_zone_limit)
1150 /* Lookup connection and confirm if unconfirmed. */
1151 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
1152 const struct ovs_conntrack_info *info,
1153 struct sk_buff *skb)
1155 enum ip_conntrack_info ctinfo;
1159 err = __ovs_ct_lookup(net, key, info, skb);
1163 /* The connection could be invalid, in which case this is a no-op.*/
1164 ct = nf_ct_get(skb, &ctinfo);
1168 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1169 if (static_branch_unlikely(&ovs_ct_limit_enabled)) {
1170 if (!nf_ct_is_confirmed(ct)) {
1171 err = ovs_ct_check_limit(net, info,
1172 &ct->tuplehash[IP_CT_DIR_ORIGINAL].tuple);
1174 net_warn_ratelimited("openvswitch: zone: %u "
1175 "exceeds conntrack limit\n",
1183 /* Set the conntrack event mask if given. NEW and DELETE events have
1184 * their own groups, but the NFNLGRP_CONNTRACK_UPDATE group listener
1185 * typically would receive many kinds of updates. Setting the event
1186 * mask allows those events to be filtered. The set event mask will
1187 * remain in effect for the lifetime of the connection unless changed
1188 * by a further CT action with both the commit flag and the eventmask
1190 if (info->have_eventmask) {
1191 struct nf_conntrack_ecache *cache = nf_ct_ecache_find(ct);
1194 cache->ctmask = info->eventmask;
1197 /* Apply changes before confirming the connection so that the initial
1198 * conntrack NEW netlink event carries the values given in the CT
1201 if (info->mark.mask) {
1202 err = ovs_ct_set_mark(ct, key, info->mark.value,
1207 if (!nf_ct_is_confirmed(ct)) {
1208 err = ovs_ct_init_labels(ct, key, &info->labels.value,
1209 &info->labels.mask);
1212 } else if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1213 labels_nonzero(&info->labels.mask)) {
1214 err = ovs_ct_set_labels(ct, key, &info->labels.value,
1215 &info->labels.mask);
1219 /* This will take care of sending queued events even if the connection
1220 * is already confirmed.
1222 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
1228 /* Trim the skb to the length specified by the IP/IPv6 header,
1229 * removing any trailing lower-layer padding. This prepares the skb
1230 * for higher-layer processing that assumes skb->len excludes padding
1231 * (such as nf_ip_checksum). The caller needs to pull the skb to the
1232 * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
1234 static int ovs_skb_network_trim(struct sk_buff *skb)
1239 switch (skb->protocol) {
1240 case htons(ETH_P_IP):
1241 len = ntohs(ip_hdr(skb)->tot_len);
1243 case htons(ETH_P_IPV6):
1244 len = sizeof(struct ipv6hdr)
1245 + ntohs(ipv6_hdr(skb)->payload_len);
1251 err = pskb_trim_rcsum(skb, len);
1258 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
1259 * value if 'skb' is freed.
1261 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
1262 struct sw_flow_key *key,
1263 const struct ovs_conntrack_info *info)
1268 /* The conntrack module expects to be working at L3. */
1269 nh_ofs = skb_network_offset(skb);
1270 skb_pull_rcsum(skb, nh_ofs);
1272 err = ovs_skb_network_trim(skb);
1276 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
1277 err = handle_fragments(net, key, info->zone.id, skb);
1283 err = ovs_ct_commit(net, key, info, skb);
1285 err = ovs_ct_lookup(net, key, info, skb);
1287 skb_push(skb, nh_ofs);
1288 skb_postpush_rcsum(skb, skb->data, nh_ofs);
1294 int ovs_ct_clear(struct sk_buff *skb, struct sw_flow_key *key)
1296 if (skb_nfct(skb)) {
1297 nf_conntrack_put(skb_nfct(skb));
1298 nf_ct_set(skb, NULL, IP_CT_UNTRACKED);
1299 ovs_ct_fill_key(skb, key);
1305 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
1306 const struct sw_flow_key *key, bool log)
1308 struct nf_conntrack_helper *helper;
1309 struct nf_conn_help *help;
1312 helper = nf_conntrack_helper_try_module_get(name, info->family,
1315 OVS_NLERR(log, "Unknown helper \"%s\"", name);
1319 help = nf_ct_helper_ext_add(info->ct, GFP_KERNEL);
1321 nf_conntrack_helper_put(helper);
1325 #if IS_ENABLED(CONFIG_NF_NAT)
1327 ret = nf_nat_helper_try_module_get(name, info->family,
1330 nf_conntrack_helper_put(helper);
1331 OVS_NLERR(log, "Failed to load \"%s\" NAT helper, error: %d",
1337 rcu_assign_pointer(help->helper, helper);
1338 info->helper = helper;
1342 #if IS_ENABLED(CONFIG_NF_NAT)
1343 static int parse_nat(const struct nlattr *attr,
1344 struct ovs_conntrack_info *info, bool log)
1348 bool have_ip_max = false;
1349 bool have_proto_max = false;
1350 bool ip_vers = (info->family == NFPROTO_IPV6);
1352 nla_for_each_nested(a, attr, rem) {
1353 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
1354 [OVS_NAT_ATTR_SRC] = {0, 0},
1355 [OVS_NAT_ATTR_DST] = {0, 0},
1356 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
1357 sizeof(struct in6_addr)},
1358 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
1359 sizeof(struct in6_addr)},
1360 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
1361 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
1362 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
1363 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
1364 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
1366 int type = nla_type(a);
1368 if (type > OVS_NAT_ATTR_MAX) {
1369 OVS_NLERR(log, "Unknown NAT attribute (type=%d, max=%d)",
1370 type, OVS_NAT_ATTR_MAX);
1374 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
1375 OVS_NLERR(log, "NAT attribute type %d has unexpected length (%d != %d)",
1377 ovs_nat_attr_lens[type][ip_vers]);
1382 case OVS_NAT_ATTR_SRC:
1383 case OVS_NAT_ATTR_DST:
1385 OVS_NLERR(log, "Only one type of NAT may be specified");
1388 info->nat |= OVS_CT_NAT;
1389 info->nat |= ((type == OVS_NAT_ATTR_SRC)
1390 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
1393 case OVS_NAT_ATTR_IP_MIN:
1394 nla_memcpy(&info->range.min_addr, a,
1395 sizeof(info->range.min_addr));
1396 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1399 case OVS_NAT_ATTR_IP_MAX:
1401 nla_memcpy(&info->range.max_addr, a,
1402 sizeof(info->range.max_addr));
1403 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1406 case OVS_NAT_ATTR_PROTO_MIN:
1407 info->range.min_proto.all = htons(nla_get_u16(a));
1408 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1411 case OVS_NAT_ATTR_PROTO_MAX:
1412 have_proto_max = true;
1413 info->range.max_proto.all = htons(nla_get_u16(a));
1414 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1417 case OVS_NAT_ATTR_PERSISTENT:
1418 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1421 case OVS_NAT_ATTR_PROTO_HASH:
1422 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1425 case OVS_NAT_ATTR_PROTO_RANDOM:
1426 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1430 OVS_NLERR(log, "Unknown nat attribute (%d)", type);
1436 OVS_NLERR(log, "NAT attribute has %d unknown bytes", rem);
1440 /* Do not allow flags if no type is given. */
1441 if (info->range.flags) {
1443 "NAT flags may be given only when NAT range (SRC or DST) is also specified."
1447 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1448 } else if (!info->commit) {
1450 "NAT attributes may be specified only when CT COMMIT flag is also specified."
1454 /* Allow missing IP_MAX. */
1455 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1456 memcpy(&info->range.max_addr, &info->range.min_addr,
1457 sizeof(info->range.max_addr));
1459 /* Allow missing PROTO_MAX. */
1460 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1462 info->range.max_proto.all = info->range.min_proto.all;
1468 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1469 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
1470 [OVS_CT_ATTR_FORCE_COMMIT] = { .minlen = 0, .maxlen = 0 },
1471 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1472 .maxlen = sizeof(u16) },
1473 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1474 .maxlen = sizeof(struct md_mark) },
1475 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1476 .maxlen = sizeof(struct md_labels) },
1477 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
1478 .maxlen = NF_CT_HELPER_NAME_LEN },
1479 #if IS_ENABLED(CONFIG_NF_NAT)
1480 /* NAT length is checked when parsing the nested attributes. */
1481 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1483 [OVS_CT_ATTR_EVENTMASK] = { .minlen = sizeof(u32),
1484 .maxlen = sizeof(u32) },
1485 [OVS_CT_ATTR_TIMEOUT] = { .minlen = 1,
1486 .maxlen = CTNL_TIMEOUT_NAME_MAX },
1489 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1490 const char **helper, bool log)
1495 nla_for_each_nested(a, attr, rem) {
1496 int type = nla_type(a);
1500 if (type > OVS_CT_ATTR_MAX) {
1502 "Unknown conntrack attr (type=%d, max=%d)",
1503 type, OVS_CT_ATTR_MAX);
1507 maxlen = ovs_ct_attr_lens[type].maxlen;
1508 minlen = ovs_ct_attr_lens[type].minlen;
1509 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1511 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1512 type, nla_len(a), maxlen);
1517 case OVS_CT_ATTR_FORCE_COMMIT:
1520 case OVS_CT_ATTR_COMMIT:
1521 info->commit = true;
1523 #ifdef CONFIG_NF_CONNTRACK_ZONES
1524 case OVS_CT_ATTR_ZONE:
1525 info->zone.id = nla_get_u16(a);
1528 #ifdef CONFIG_NF_CONNTRACK_MARK
1529 case OVS_CT_ATTR_MARK: {
1530 struct md_mark *mark = nla_data(a);
1533 OVS_NLERR(log, "ct_mark mask cannot be 0");
1540 #ifdef CONFIG_NF_CONNTRACK_LABELS
1541 case OVS_CT_ATTR_LABELS: {
1542 struct md_labels *labels = nla_data(a);
1544 if (!labels_nonzero(&labels->mask)) {
1545 OVS_NLERR(log, "ct_labels mask cannot be 0");
1548 info->labels = *labels;
1552 case OVS_CT_ATTR_HELPER:
1553 *helper = nla_data(a);
1554 if (!memchr(*helper, '\0', nla_len(a))) {
1555 OVS_NLERR(log, "Invalid conntrack helper");
1559 #if IS_ENABLED(CONFIG_NF_NAT)
1560 case OVS_CT_ATTR_NAT: {
1561 int err = parse_nat(a, info, log);
1568 case OVS_CT_ATTR_EVENTMASK:
1569 info->have_eventmask = true;
1570 info->eventmask = nla_get_u32(a);
1572 #ifdef CONFIG_NF_CONNTRACK_TIMEOUT
1573 case OVS_CT_ATTR_TIMEOUT:
1574 memcpy(info->timeout, nla_data(a), nla_len(a));
1575 if (!memchr(info->timeout, '\0', nla_len(a))) {
1576 OVS_NLERR(log, "Invalid conntrack helper");
1583 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1589 #ifdef CONFIG_NF_CONNTRACK_MARK
1590 if (!info->commit && info->mark.mask) {
1592 "Setting conntrack mark requires 'commit' flag.");
1596 #ifdef CONFIG_NF_CONNTRACK_LABELS
1597 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1599 "Setting conntrack labels requires 'commit' flag.");
1604 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1611 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1613 if (attr == OVS_KEY_ATTR_CT_STATE)
1615 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1616 attr == OVS_KEY_ATTR_CT_ZONE)
1618 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1619 attr == OVS_KEY_ATTR_CT_MARK)
1621 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1622 attr == OVS_KEY_ATTR_CT_LABELS) {
1623 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1625 return ovs_net->xt_label;
1631 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1632 const struct sw_flow_key *key,
1633 struct sw_flow_actions **sfa, bool log)
1635 struct ovs_conntrack_info ct_info;
1636 const char *helper = NULL;
1640 family = key_to_nfproto(key);
1641 if (family == NFPROTO_UNSPEC) {
1642 OVS_NLERR(log, "ct family unspecified");
1646 memset(&ct_info, 0, sizeof(ct_info));
1647 ct_info.family = family;
1649 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1650 NF_CT_DEFAULT_ZONE_DIR, 0);
1652 err = parse_ct(attr, &ct_info, &helper, log);
1656 /* Set up template for tracking connections in specific zones. */
1657 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1659 OVS_NLERR(log, "Failed to allocate conntrack template");
1663 if (ct_info.timeout[0]) {
1664 if (nf_ct_set_timeout(net, ct_info.ct, family, key->ip.proto,
1666 pr_info_ratelimited("Failed to associated timeout "
1667 "policy `%s'\n", ct_info.timeout);
1671 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1676 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1677 sizeof(ct_info), log);
1681 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1682 nf_conntrack_get(&ct_info.ct->ct_general);
1685 __ovs_ct_free_action(&ct_info);
1689 #if IS_ENABLED(CONFIG_NF_NAT)
1690 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1691 struct sk_buff *skb)
1693 struct nlattr *start;
1695 start = nla_nest_start_noflag(skb, OVS_CT_ATTR_NAT);
1699 if (info->nat & OVS_CT_SRC_NAT) {
1700 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1702 } else if (info->nat & OVS_CT_DST_NAT) {
1703 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1709 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1710 if (IS_ENABLED(CONFIG_NF_NAT) &&
1711 info->family == NFPROTO_IPV4) {
1712 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1713 info->range.min_addr.ip) ||
1714 (info->range.max_addr.ip
1715 != info->range.min_addr.ip &&
1716 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1717 info->range.max_addr.ip))))
1719 } else if (IS_ENABLED(CONFIG_IPV6) &&
1720 info->family == NFPROTO_IPV6) {
1721 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1722 &info->range.min_addr.in6) ||
1723 (memcmp(&info->range.max_addr.in6,
1724 &info->range.min_addr.in6,
1725 sizeof(info->range.max_addr.in6)) &&
1726 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1727 &info->range.max_addr.in6))))
1733 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1734 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1735 ntohs(info->range.min_proto.all)) ||
1736 (info->range.max_proto.all != info->range.min_proto.all &&
1737 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1738 ntohs(info->range.max_proto.all)))))
1741 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1742 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1744 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1745 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1747 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1748 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1751 nla_nest_end(skb, start);
1757 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1758 struct sk_buff *skb)
1760 struct nlattr *start;
1762 start = nla_nest_start_noflag(skb, OVS_ACTION_ATTR_CT);
1766 if (ct_info->commit && nla_put_flag(skb, ct_info->force
1767 ? OVS_CT_ATTR_FORCE_COMMIT
1768 : OVS_CT_ATTR_COMMIT))
1770 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1771 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1773 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1774 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1777 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1778 labels_nonzero(&ct_info->labels.mask) &&
1779 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1782 if (ct_info->helper) {
1783 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1784 ct_info->helper->name))
1787 if (ct_info->have_eventmask &&
1788 nla_put_u32(skb, OVS_CT_ATTR_EVENTMASK, ct_info->eventmask))
1790 if (ct_info->timeout[0]) {
1791 if (nla_put_string(skb, OVS_CT_ATTR_TIMEOUT, ct_info->timeout))
1795 #if IS_ENABLED(CONFIG_NF_NAT)
1796 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1799 nla_nest_end(skb, start);
1804 void ovs_ct_free_action(const struct nlattr *a)
1806 struct ovs_conntrack_info *ct_info = nla_data(a);
1808 __ovs_ct_free_action(ct_info);
1811 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1813 if (ct_info->helper) {
1814 #if IS_ENABLED(CONFIG_NF_NAT)
1816 nf_nat_helper_put(ct_info->helper);
1818 nf_conntrack_helper_put(ct_info->helper);
1821 if (ct_info->timeout[0])
1822 nf_ct_destroy_timeout(ct_info->ct);
1823 nf_ct_tmpl_free(ct_info->ct);
1827 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
1828 static int ovs_ct_limit_init(struct net *net, struct ovs_net *ovs_net)
1832 ovs_net->ct_limit_info = kmalloc(sizeof(*ovs_net->ct_limit_info),
1834 if (!ovs_net->ct_limit_info)
1837 ovs_net->ct_limit_info->default_limit = OVS_CT_LIMIT_DEFAULT;
1838 ovs_net->ct_limit_info->limits =
1839 kmalloc_array(CT_LIMIT_HASH_BUCKETS, sizeof(struct hlist_head),
1841 if (!ovs_net->ct_limit_info->limits) {
1842 kfree(ovs_net->ct_limit_info);
1846 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; i++)
1847 INIT_HLIST_HEAD(&ovs_net->ct_limit_info->limits[i]);
1849 ovs_net->ct_limit_info->data =
1850 nf_conncount_init(net, NFPROTO_INET, sizeof(u32));
1852 if (IS_ERR(ovs_net->ct_limit_info->data)) {
1853 err = PTR_ERR(ovs_net->ct_limit_info->data);
1854 kfree(ovs_net->ct_limit_info->limits);
1855 kfree(ovs_net->ct_limit_info);
1856 pr_err("openvswitch: failed to init nf_conncount %d\n", err);
1862 static void ovs_ct_limit_exit(struct net *net, struct ovs_net *ovs_net)
1864 const struct ovs_ct_limit_info *info = ovs_net->ct_limit_info;
1867 nf_conncount_destroy(net, NFPROTO_INET, info->data);
1868 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; ++i) {
1869 struct hlist_head *head = &info->limits[i];
1870 struct ovs_ct_limit *ct_limit;
1872 hlist_for_each_entry_rcu(ct_limit, head, hlist_node)
1873 kfree_rcu(ct_limit, rcu);
1875 kfree(ovs_net->ct_limit_info->limits);
1876 kfree(ovs_net->ct_limit_info);
1879 static struct sk_buff *
1880 ovs_ct_limit_cmd_reply_start(struct genl_info *info, u8 cmd,
1881 struct ovs_header **ovs_reply_header)
1883 struct ovs_header *ovs_header = info->userhdr;
1884 struct sk_buff *skb;
1886 skb = genlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1888 return ERR_PTR(-ENOMEM);
1890 *ovs_reply_header = genlmsg_put(skb, info->snd_portid,
1892 &dp_ct_limit_genl_family, 0, cmd);
1894 if (!*ovs_reply_header) {
1896 return ERR_PTR(-EMSGSIZE);
1898 (*ovs_reply_header)->dp_ifindex = ovs_header->dp_ifindex;
1903 static bool check_zone_id(int zone_id, u16 *pzone)
1905 if (zone_id >= 0 && zone_id <= 65535) {
1906 *pzone = (u16)zone_id;
1912 static int ovs_ct_limit_set_zone_limit(struct nlattr *nla_zone_limit,
1913 struct ovs_ct_limit_info *info)
1915 struct ovs_zone_limit *zone_limit;
1919 rem = NLA_ALIGN(nla_len(nla_zone_limit));
1920 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
1922 while (rem >= sizeof(*zone_limit)) {
1923 if (unlikely(zone_limit->zone_id ==
1924 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
1926 info->default_limit = zone_limit->limit;
1928 } else if (unlikely(!check_zone_id(
1929 zone_limit->zone_id, &zone))) {
1930 OVS_NLERR(true, "zone id is out of range");
1932 struct ovs_ct_limit *ct_limit;
1934 ct_limit = kmalloc(sizeof(*ct_limit), GFP_KERNEL);
1938 ct_limit->zone = zone;
1939 ct_limit->limit = zone_limit->limit;
1942 ct_limit_set(info, ct_limit);
1945 rem -= NLA_ALIGN(sizeof(*zone_limit));
1946 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
1947 NLA_ALIGN(sizeof(*zone_limit)));
1951 OVS_NLERR(true, "set zone limit has %d unknown bytes", rem);
1956 static int ovs_ct_limit_del_zone_limit(struct nlattr *nla_zone_limit,
1957 struct ovs_ct_limit_info *info)
1959 struct ovs_zone_limit *zone_limit;
1963 rem = NLA_ALIGN(nla_len(nla_zone_limit));
1964 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
1966 while (rem >= sizeof(*zone_limit)) {
1967 if (unlikely(zone_limit->zone_id ==
1968 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
1970 info->default_limit = OVS_CT_LIMIT_DEFAULT;
1972 } else if (unlikely(!check_zone_id(
1973 zone_limit->zone_id, &zone))) {
1974 OVS_NLERR(true, "zone id is out of range");
1977 ct_limit_del(info, zone);
1980 rem -= NLA_ALIGN(sizeof(*zone_limit));
1981 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
1982 NLA_ALIGN(sizeof(*zone_limit)));
1986 OVS_NLERR(true, "del zone limit has %d unknown bytes", rem);
1991 static int ovs_ct_limit_get_default_limit(struct ovs_ct_limit_info *info,
1992 struct sk_buff *reply)
1994 struct ovs_zone_limit zone_limit;
1997 zone_limit.zone_id = OVS_ZONE_LIMIT_DEFAULT_ZONE;
1998 zone_limit.limit = info->default_limit;
1999 err = nla_put_nohdr(reply, sizeof(zone_limit), &zone_limit);
2006 static int __ovs_ct_limit_get_zone_limit(struct net *net,
2007 struct nf_conncount_data *data,
2008 u16 zone_id, u32 limit,
2009 struct sk_buff *reply)
2011 struct nf_conntrack_zone ct_zone;
2012 struct ovs_zone_limit zone_limit;
2013 u32 conncount_key = zone_id;
2015 zone_limit.zone_id = zone_id;
2016 zone_limit.limit = limit;
2017 nf_ct_zone_init(&ct_zone, zone_id, NF_CT_DEFAULT_ZONE_DIR, 0);
2019 zone_limit.count = nf_conncount_count(net, data, &conncount_key, NULL,
2021 return nla_put_nohdr(reply, sizeof(zone_limit), &zone_limit);
2024 static int ovs_ct_limit_get_zone_limit(struct net *net,
2025 struct nlattr *nla_zone_limit,
2026 struct ovs_ct_limit_info *info,
2027 struct sk_buff *reply)
2029 struct ovs_zone_limit *zone_limit;
2034 rem = NLA_ALIGN(nla_len(nla_zone_limit));
2035 zone_limit = (struct ovs_zone_limit *)nla_data(nla_zone_limit);
2037 while (rem >= sizeof(*zone_limit)) {
2038 if (unlikely(zone_limit->zone_id ==
2039 OVS_ZONE_LIMIT_DEFAULT_ZONE)) {
2040 err = ovs_ct_limit_get_default_limit(info, reply);
2043 } else if (unlikely(!check_zone_id(zone_limit->zone_id,
2045 OVS_NLERR(true, "zone id is out of range");
2048 limit = ct_limit_get(info, zone);
2051 err = __ovs_ct_limit_get_zone_limit(
2052 net, info->data, zone, limit, reply);
2056 rem -= NLA_ALIGN(sizeof(*zone_limit));
2057 zone_limit = (struct ovs_zone_limit *)((u8 *)zone_limit +
2058 NLA_ALIGN(sizeof(*zone_limit)));
2062 OVS_NLERR(true, "get zone limit has %d unknown bytes", rem);
2067 static int ovs_ct_limit_get_all_zone_limit(struct net *net,
2068 struct ovs_ct_limit_info *info,
2069 struct sk_buff *reply)
2071 struct ovs_ct_limit *ct_limit;
2072 struct hlist_head *head;
2075 err = ovs_ct_limit_get_default_limit(info, reply);
2080 for (i = 0; i < CT_LIMIT_HASH_BUCKETS; ++i) {
2081 head = &info->limits[i];
2082 hlist_for_each_entry_rcu(ct_limit, head, hlist_node) {
2083 err = __ovs_ct_limit_get_zone_limit(net, info->data,
2084 ct_limit->zone, ct_limit->limit, reply);
2095 static int ovs_ct_limit_cmd_set(struct sk_buff *skb, struct genl_info *info)
2097 struct nlattr **a = info->attrs;
2098 struct sk_buff *reply;
2099 struct ovs_header *ovs_reply_header;
2100 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
2101 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2104 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_SET,
2107 return PTR_ERR(reply);
2109 if (!a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2114 err = ovs_ct_limit_set_zone_limit(a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT],
2119 static_branch_enable(&ovs_ct_limit_enabled);
2121 genlmsg_end(reply, ovs_reply_header);
2122 return genlmsg_reply(reply, info);
2129 static int ovs_ct_limit_cmd_del(struct sk_buff *skb, struct genl_info *info)
2131 struct nlattr **a = info->attrs;
2132 struct sk_buff *reply;
2133 struct ovs_header *ovs_reply_header;
2134 struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id);
2135 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2138 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_DEL,
2141 return PTR_ERR(reply);
2143 if (!a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2148 err = ovs_ct_limit_del_zone_limit(a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT],
2153 genlmsg_end(reply, ovs_reply_header);
2154 return genlmsg_reply(reply, info);
2161 static int ovs_ct_limit_cmd_get(struct sk_buff *skb, struct genl_info *info)
2163 struct nlattr **a = info->attrs;
2164 struct nlattr *nla_reply;
2165 struct sk_buff *reply;
2166 struct ovs_header *ovs_reply_header;
2167 struct net *net = sock_net(skb->sk);
2168 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2169 struct ovs_ct_limit_info *ct_limit_info = ovs_net->ct_limit_info;
2172 reply = ovs_ct_limit_cmd_reply_start(info, OVS_CT_LIMIT_CMD_GET,
2175 return PTR_ERR(reply);
2177 nla_reply = nla_nest_start_noflag(reply, OVS_CT_LIMIT_ATTR_ZONE_LIMIT);
2183 if (a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT]) {
2184 err = ovs_ct_limit_get_zone_limit(
2185 net, a[OVS_CT_LIMIT_ATTR_ZONE_LIMIT], ct_limit_info,
2190 err = ovs_ct_limit_get_all_zone_limit(net, ct_limit_info,
2196 nla_nest_end(reply, nla_reply);
2197 genlmsg_end(reply, ovs_reply_header);
2198 return genlmsg_reply(reply, info);
2205 static struct genl_ops ct_limit_genl_ops[] = {
2206 { .cmd = OVS_CT_LIMIT_CMD_SET,
2207 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2208 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN
2210 .doit = ovs_ct_limit_cmd_set,
2212 { .cmd = OVS_CT_LIMIT_CMD_DEL,
2213 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2214 .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN
2216 .doit = ovs_ct_limit_cmd_del,
2218 { .cmd = OVS_CT_LIMIT_CMD_GET,
2219 .validate = GENL_DONT_VALIDATE_STRICT | GENL_DONT_VALIDATE_DUMP,
2220 .flags = 0, /* OK for unprivileged users. */
2221 .doit = ovs_ct_limit_cmd_get,
2225 static const struct genl_multicast_group ovs_ct_limit_multicast_group = {
2226 .name = OVS_CT_LIMIT_MCGROUP,
2229 struct genl_family dp_ct_limit_genl_family __ro_after_init = {
2230 .hdrsize = sizeof(struct ovs_header),
2231 .name = OVS_CT_LIMIT_FAMILY,
2232 .version = OVS_CT_LIMIT_VERSION,
2233 .maxattr = OVS_CT_LIMIT_ATTR_MAX,
2234 .policy = ct_limit_policy,
2236 .parallel_ops = true,
2237 .ops = ct_limit_genl_ops,
2238 .n_ops = ARRAY_SIZE(ct_limit_genl_ops),
2239 .mcgrps = &ovs_ct_limit_multicast_group,
2241 .module = THIS_MODULE,
2245 int ovs_ct_init(struct net *net)
2247 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
2248 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2250 if (nf_connlabels_get(net, n_bits - 1)) {
2251 ovs_net->xt_label = false;
2252 OVS_NLERR(true, "Failed to set connlabel length");
2254 ovs_net->xt_label = true;
2257 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
2258 return ovs_ct_limit_init(net, ovs_net);
2264 void ovs_ct_exit(struct net *net)
2266 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2268 #if IS_ENABLED(CONFIG_NETFILTER_CONNCOUNT)
2269 ovs_ct_limit_exit(net, ovs_net);
2272 if (ovs_net->xt_label)
2273 nf_connlabels_put(net);