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
= sizeof(cl
->bits
);
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
|| sizeof(cl
->bits
) < 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
);
442 if (l3proto
->get_l4proto(skb
, skb_network_offset(skb
), &dataoff
,
444 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
447 l4proto
= __nf_ct_l4proto_find(l3num
, protonum
);
448 if (!nf_ct_get_tuple(skb
, skb_network_offset(skb
), dataoff
, l3num
,
449 protonum
, net
, &tuple
, l3proto
, l4proto
)) {
450 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
454 /* look for tuple match */
455 h
= nf_conntrack_find_get(net
, zone
, &tuple
);
457 return NULL
; /* Not found. */
459 ct
= nf_ct_tuplehash_to_ctrack(h
);
461 ctinfo
= ovs_ct_get_info(h
);
462 if (ctinfo
== IP_CT_NEW
) {
463 /* This should not happen. */
464 WARN_ONCE(1, "ovs_ct_find_existing: new packet for %p\n", ct
);
466 skb
->nfct
= &ct
->ct_general
;
467 skb
->nfctinfo
= ctinfo
;
471 /* Determine whether skb->nfct is equal to the result of conntrack lookup. */
472 static bool skb_nfct_cached(struct net
*net
,
473 const struct sw_flow_key
*key
,
474 const struct ovs_conntrack_info
*info
,
477 enum ip_conntrack_info ctinfo
;
480 ct
= nf_ct_get(skb
, &ctinfo
);
481 /* If no ct, check if we have evidence that an existing conntrack entry
482 * might be found for this skb. This happens when we lose a skb->nfct
483 * due to an upcall. If the connection was not confirmed, it is not
484 * cached and needs to be run through conntrack again.
486 if (!ct
&& key
->ct
.state
& OVS_CS_F_TRACKED
&&
487 !(key
->ct
.state
& OVS_CS_F_INVALID
) &&
488 key
->ct
.zone
== info
->zone
.id
)
489 ct
= ovs_ct_find_existing(net
, &info
->zone
, info
->family
, skb
);
492 if (!net_eq(net
, read_pnet(&ct
->ct_net
)))
494 if (!nf_ct_zone_equal_any(info
->ct
, nf_ct_zone(ct
)))
497 struct nf_conn_help
*help
;
499 help
= nf_ct_ext_find(ct
, NF_CT_EXT_HELPER
);
500 if (help
&& rcu_access_pointer(help
->helper
) != info
->helper
)
507 #ifdef CONFIG_NF_NAT_NEEDED
508 /* Modelled after nf_nat_ipv[46]_fn().
509 * range is only used for new, uninitialized NAT state.
510 * Returns either NF_ACCEPT or NF_DROP.
512 static int ovs_ct_nat_execute(struct sk_buff
*skb
, struct nf_conn
*ct
,
513 enum ip_conntrack_info ctinfo
,
514 const struct nf_nat_range
*range
,
515 enum nf_nat_manip_type maniptype
)
517 int hooknum
, nh_off
, err
= NF_ACCEPT
;
519 nh_off
= skb_network_offset(skb
);
520 skb_pull(skb
, nh_off
);
522 /* See HOOK2MANIP(). */
523 if (maniptype
== NF_NAT_MANIP_SRC
)
524 hooknum
= NF_INET_LOCAL_IN
; /* Source NAT */
526 hooknum
= NF_INET_LOCAL_OUT
; /* Destination NAT */
530 case IP_CT_RELATED_REPLY
:
531 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
532 skb
->protocol
== htons(ETH_P_IP
) &&
533 ip_hdr(skb
)->protocol
== IPPROTO_ICMP
) {
534 if (!nf_nat_icmp_reply_translation(skb
, ct
, ctinfo
,
538 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
539 skb
->protocol
== htons(ETH_P_IPV6
)) {
541 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
542 int hdrlen
= ipv6_skip_exthdr(skb
,
543 sizeof(struct ipv6hdr
),
544 &nexthdr
, &frag_off
);
546 if (hdrlen
>= 0 && nexthdr
== IPPROTO_ICMPV6
) {
547 if (!nf_nat_icmpv6_reply_translation(skb
, ct
,
555 /* Non-ICMP, fall thru to initialize if needed. */
557 /* Seen it before? This can happen for loopback, retrans,
560 if (!nf_nat_initialized(ct
, maniptype
)) {
561 /* Initialize according to the NAT action. */
562 err
= (range
&& range
->flags
& NF_NAT_RANGE_MAP_IPS
)
563 /* Action is set up to establish a new
566 ? nf_nat_setup_info(ct
, range
, maniptype
)
567 : nf_nat_alloc_null_binding(ct
, hooknum
);
568 if (err
!= NF_ACCEPT
)
573 case IP_CT_ESTABLISHED
:
574 case IP_CT_ESTABLISHED_REPLY
:
582 err
= nf_nat_packet(ct
, ctinfo
, hooknum
, skb
);
584 skb_push(skb
, nh_off
);
589 static void ovs_nat_update_key(struct sw_flow_key
*key
,
590 const struct sk_buff
*skb
,
591 enum nf_nat_manip_type maniptype
)
593 if (maniptype
== NF_NAT_MANIP_SRC
) {
596 key
->ct
.state
|= OVS_CS_F_SRC_NAT
;
597 if (key
->eth
.type
== htons(ETH_P_IP
))
598 key
->ipv4
.addr
.src
= ip_hdr(skb
)->saddr
;
599 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
600 memcpy(&key
->ipv6
.addr
.src
, &ipv6_hdr(skb
)->saddr
,
601 sizeof(key
->ipv6
.addr
.src
));
605 if (key
->ip
.proto
== IPPROTO_UDP
)
606 src
= udp_hdr(skb
)->source
;
607 else if (key
->ip
.proto
== IPPROTO_TCP
)
608 src
= tcp_hdr(skb
)->source
;
609 else if (key
->ip
.proto
== IPPROTO_SCTP
)
610 src
= sctp_hdr(skb
)->source
;
618 key
->ct
.state
|= OVS_CS_F_DST_NAT
;
619 if (key
->eth
.type
== htons(ETH_P_IP
))
620 key
->ipv4
.addr
.dst
= ip_hdr(skb
)->daddr
;
621 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
622 memcpy(&key
->ipv6
.addr
.dst
, &ipv6_hdr(skb
)->daddr
,
623 sizeof(key
->ipv6
.addr
.dst
));
627 if (key
->ip
.proto
== IPPROTO_UDP
)
628 dst
= udp_hdr(skb
)->dest
;
629 else if (key
->ip
.proto
== IPPROTO_TCP
)
630 dst
= tcp_hdr(skb
)->dest
;
631 else if (key
->ip
.proto
== IPPROTO_SCTP
)
632 dst
= sctp_hdr(skb
)->dest
;
640 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
641 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
642 const struct ovs_conntrack_info
*info
,
643 struct sk_buff
*skb
, struct nf_conn
*ct
,
644 enum ip_conntrack_info ctinfo
)
646 enum nf_nat_manip_type maniptype
;
649 if (nf_ct_is_untracked(ct
)) {
650 /* A NAT action may only be performed on tracked packets. */
654 /* Add NAT extension if not confirmed yet. */
655 if (!nf_ct_is_confirmed(ct
) && !nf_ct_nat_ext_add(ct
))
656 return NF_ACCEPT
; /* Can't NAT. */
658 /* Determine NAT type.
659 * Check if the NAT type can be deduced from the tracked connection.
660 * Make sure new expected connections (IP_CT_RELATED) are NATted only
663 if (info
->nat
& OVS_CT_NAT
&& ctinfo
!= IP_CT_NEW
&&
664 ct
->status
& IPS_NAT_MASK
&&
665 (ctinfo
!= IP_CT_RELATED
|| info
->commit
)) {
666 /* NAT an established or related connection like before. */
667 if (CTINFO2DIR(ctinfo
) == IP_CT_DIR_REPLY
)
668 /* This is the REPLY direction for a connection
669 * for which NAT was applied in the forward
670 * direction. Do the reverse NAT.
672 maniptype
= ct
->status
& IPS_SRC_NAT
673 ? NF_NAT_MANIP_DST
: NF_NAT_MANIP_SRC
;
675 maniptype
= ct
->status
& IPS_SRC_NAT
676 ? NF_NAT_MANIP_SRC
: NF_NAT_MANIP_DST
;
677 } else if (info
->nat
& OVS_CT_SRC_NAT
) {
678 maniptype
= NF_NAT_MANIP_SRC
;
679 } else if (info
->nat
& OVS_CT_DST_NAT
) {
680 maniptype
= NF_NAT_MANIP_DST
;
682 return NF_ACCEPT
; /* Connection is not NATed. */
684 err
= ovs_ct_nat_execute(skb
, ct
, ctinfo
, &info
->range
, maniptype
);
686 /* Mark NAT done if successful and update the flow key. */
687 if (err
== NF_ACCEPT
)
688 ovs_nat_update_key(key
, skb
, maniptype
);
692 #else /* !CONFIG_NF_NAT_NEEDED */
693 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
694 const struct ovs_conntrack_info
*info
,
695 struct sk_buff
*skb
, struct nf_conn
*ct
,
696 enum ip_conntrack_info ctinfo
)
702 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
703 * not done already. Update key with new CT state after passing the packet
705 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
706 * set to NULL and 0 will be returned.
708 static int __ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
709 const struct ovs_conntrack_info
*info
,
712 /* If we are recirculating packets to match on conntrack fields and
713 * committing with a separate conntrack action, then we don't need to
714 * actually run the packet through conntrack twice unless it's for a
717 bool cached
= skb_nfct_cached(net
, key
, info
, skb
);
718 enum ip_conntrack_info ctinfo
;
722 struct nf_conn
*tmpl
= info
->ct
;
725 /* Associate skb with specified zone. */
728 nf_conntrack_put(skb
->nfct
);
729 nf_conntrack_get(&tmpl
->ct_general
);
730 skb
->nfct
= &tmpl
->ct_general
;
731 skb
->nfctinfo
= IP_CT_NEW
;
734 /* Repeat if requested, see nf_iterate(). */
736 err
= nf_conntrack_in(net
, info
->family
,
737 NF_INET_PRE_ROUTING
, skb
);
738 } while (err
== NF_REPEAT
);
740 if (err
!= NF_ACCEPT
)
743 /* Clear CT state NAT flags to mark that we have not yet done
744 * NAT after the nf_conntrack_in() call. We can actually clear
745 * the whole state, as it will be re-initialized below.
749 /* Update the key, but keep the NAT flags. */
750 ovs_ct_update_key(skb
, info
, key
, true, true);
753 ct
= nf_ct_get(skb
, &ctinfo
);
755 /* Packets starting a new connection must be NATted before the
756 * helper, so that the helper knows about the NAT. We enforce
757 * this by delaying both NAT and helper calls for unconfirmed
758 * connections until the committing CT action. For later
759 * packets NAT and Helper may be called in either order.
761 * NAT will be done only if the CT action has NAT, and only
762 * once per packet (per zone), as guarded by the NAT bits in
765 if (info
->nat
&& !(key
->ct
.state
& OVS_CS_F_NAT_MASK
) &&
766 (nf_ct_is_confirmed(ct
) || info
->commit
) &&
767 ovs_ct_nat(net
, key
, info
, skb
, ct
, ctinfo
) != NF_ACCEPT
) {
771 /* Userspace may decide to perform a ct lookup without a helper
772 * specified followed by a (recirculate and) commit with one.
773 * Therefore, for unconfirmed connections which we will commit,
774 * we need to attach the helper here.
776 if (!nf_ct_is_confirmed(ct
) && info
->commit
&&
777 info
->helper
&& !nfct_help(ct
)) {
778 int err
= __nf_ct_try_assign_helper(ct
, info
->ct
,
784 /* Call the helper only if:
785 * - nf_conntrack_in() was executed above ("!cached") for a
786 * confirmed connection, or
787 * - When committing an unconfirmed connection.
789 if ((nf_ct_is_confirmed(ct
) ? !cached
: info
->commit
) &&
790 ovs_ct_helper(skb
, info
->family
) != NF_ACCEPT
) {
798 /* Lookup connection and read fields into key. */
799 static int ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
800 const struct ovs_conntrack_info
*info
,
803 struct nf_conntrack_expect
*exp
;
805 /* If we pass an expected packet through nf_conntrack_in() the
806 * expectation is typically removed, but the packet could still be
807 * lost in upcall processing. To prevent this from happening we
808 * perform an explicit expectation lookup. Expected connections are
809 * always new, and will be passed through conntrack only when they are
810 * committed, as it is OK to remove the expectation at that time.
812 exp
= ovs_ct_expect_find(net
, &info
->zone
, info
->family
, skb
);
816 /* NOTE: New connections are NATted and Helped only when
817 * committed, so we are not calling into NAT here.
819 state
= OVS_CS_F_TRACKED
| OVS_CS_F_NEW
| OVS_CS_F_RELATED
;
820 __ovs_ct_update_key(key
, state
, &info
->zone
, exp
->master
);
825 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
829 ct
= (struct nf_conn
*)skb
->nfct
;
831 nf_ct_deliver_cached_events(ct
);
837 static bool labels_nonzero(const struct ovs_key_ct_labels
*labels
)
841 for (i
= 0; i
< sizeof(*labels
); i
++)
842 if (labels
->ct_labels
[i
])
848 /* Lookup connection and confirm if unconfirmed. */
849 static int ovs_ct_commit(struct net
*net
, struct sw_flow_key
*key
,
850 const struct ovs_conntrack_info
*info
,
855 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
859 /* Apply changes before confirming the connection so that the initial
860 * conntrack NEW netlink event carries the values given in the CT
863 if (info
->mark
.mask
) {
864 err
= ovs_ct_set_mark(skb
, key
, info
->mark
.value
,
869 if (labels_nonzero(&info
->labels
.mask
)) {
870 err
= ovs_ct_set_labels(skb
, key
, &info
->labels
.value
,
875 /* This will take care of sending queued events even if the connection
876 * is already confirmed.
878 if (nf_conntrack_confirm(skb
) != NF_ACCEPT
)
884 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
885 * value if 'skb' is freed.
887 int ovs_ct_execute(struct net
*net
, struct sk_buff
*skb
,
888 struct sw_flow_key
*key
,
889 const struct ovs_conntrack_info
*info
)
894 /* The conntrack module expects to be working at L3. */
895 nh_ofs
= skb_network_offset(skb
);
896 skb_pull(skb
, nh_ofs
);
898 if (key
->ip
.frag
!= OVS_FRAG_TYPE_NONE
) {
899 err
= handle_fragments(net
, key
, info
->zone
.id
, skb
);
905 err
= ovs_ct_commit(net
, key
, info
, skb
);
907 err
= ovs_ct_lookup(net
, key
, info
, skb
);
909 skb_push(skb
, nh_ofs
);
915 static int ovs_ct_add_helper(struct ovs_conntrack_info
*info
, const char *name
,
916 const struct sw_flow_key
*key
, bool log
)
918 struct nf_conntrack_helper
*helper
;
919 struct nf_conn_help
*help
;
921 helper
= nf_conntrack_helper_try_module_get(name
, info
->family
,
924 OVS_NLERR(log
, "Unknown helper \"%s\"", name
);
928 help
= nf_ct_helper_ext_add(info
->ct
, helper
, GFP_KERNEL
);
930 module_put(helper
->me
);
934 rcu_assign_pointer(help
->helper
, helper
);
935 info
->helper
= helper
;
939 #ifdef CONFIG_NF_NAT_NEEDED
940 static int parse_nat(const struct nlattr
*attr
,
941 struct ovs_conntrack_info
*info
, bool log
)
945 bool have_ip_max
= false;
946 bool have_proto_max
= false;
947 bool ip_vers
= (info
->family
== NFPROTO_IPV6
);
949 nla_for_each_nested(a
, attr
, rem
) {
950 static const int ovs_nat_attr_lens
[OVS_NAT_ATTR_MAX
+ 1][2] = {
951 [OVS_NAT_ATTR_SRC
] = {0, 0},
952 [OVS_NAT_ATTR_DST
] = {0, 0},
953 [OVS_NAT_ATTR_IP_MIN
] = {sizeof(struct in_addr
),
954 sizeof(struct in6_addr
)},
955 [OVS_NAT_ATTR_IP_MAX
] = {sizeof(struct in_addr
),
956 sizeof(struct in6_addr
)},
957 [OVS_NAT_ATTR_PROTO_MIN
] = {sizeof(u16
), sizeof(u16
)},
958 [OVS_NAT_ATTR_PROTO_MAX
] = {sizeof(u16
), sizeof(u16
)},
959 [OVS_NAT_ATTR_PERSISTENT
] = {0, 0},
960 [OVS_NAT_ATTR_PROTO_HASH
] = {0, 0},
961 [OVS_NAT_ATTR_PROTO_RANDOM
] = {0, 0},
963 int type
= nla_type(a
);
965 if (type
> OVS_NAT_ATTR_MAX
) {
967 "Unknown NAT attribute (type=%d, max=%d).\n",
968 type
, OVS_NAT_ATTR_MAX
);
972 if (nla_len(a
) != ovs_nat_attr_lens
[type
][ip_vers
]) {
974 "NAT attribute type %d has unexpected length (%d != %d).\n",
976 ovs_nat_attr_lens
[type
][ip_vers
]);
981 case OVS_NAT_ATTR_SRC
:
982 case OVS_NAT_ATTR_DST
:
985 "Only one type of NAT may be specified.\n"
989 info
->nat
|= OVS_CT_NAT
;
990 info
->nat
|= ((type
== OVS_NAT_ATTR_SRC
)
991 ? OVS_CT_SRC_NAT
: OVS_CT_DST_NAT
);
994 case OVS_NAT_ATTR_IP_MIN
:
995 nla_memcpy(&info
->range
.min_addr
, a
,
996 sizeof(info
->range
.min_addr
));
997 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1000 case OVS_NAT_ATTR_IP_MAX
:
1002 nla_memcpy(&info
->range
.max_addr
, a
,
1003 sizeof(info
->range
.max_addr
));
1004 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1007 case OVS_NAT_ATTR_PROTO_MIN
:
1008 info
->range
.min_proto
.all
= htons(nla_get_u16(a
));
1009 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1012 case OVS_NAT_ATTR_PROTO_MAX
:
1013 have_proto_max
= true;
1014 info
->range
.max_proto
.all
= htons(nla_get_u16(a
));
1015 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1018 case OVS_NAT_ATTR_PERSISTENT
:
1019 info
->range
.flags
|= NF_NAT_RANGE_PERSISTENT
;
1022 case OVS_NAT_ATTR_PROTO_HASH
:
1023 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM
;
1026 case OVS_NAT_ATTR_PROTO_RANDOM
:
1027 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM_FULLY
;
1031 OVS_NLERR(log
, "Unknown nat attribute (%d).\n", type
);
1037 OVS_NLERR(log
, "NAT attribute has %d unknown bytes.\n", rem
);
1041 /* Do not allow flags if no type is given. */
1042 if (info
->range
.flags
) {
1044 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1048 info
->nat
= OVS_CT_NAT
; /* NAT existing connections. */
1049 } else if (!info
->commit
) {
1051 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1055 /* Allow missing IP_MAX. */
1056 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
&& !have_ip_max
) {
1057 memcpy(&info
->range
.max_addr
, &info
->range
.min_addr
,
1058 sizeof(info
->range
.max_addr
));
1060 /* Allow missing PROTO_MAX. */
1061 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1063 info
->range
.max_proto
.all
= info
->range
.min_proto
.all
;
1069 static const struct ovs_ct_len_tbl ovs_ct_attr_lens
[OVS_CT_ATTR_MAX
+ 1] = {
1070 [OVS_CT_ATTR_COMMIT
] = { .minlen
= 0, .maxlen
= 0 },
1071 [OVS_CT_ATTR_ZONE
] = { .minlen
= sizeof(u16
),
1072 .maxlen
= sizeof(u16
) },
1073 [OVS_CT_ATTR_MARK
] = { .minlen
= sizeof(struct md_mark
),
1074 .maxlen
= sizeof(struct md_mark
) },
1075 [OVS_CT_ATTR_LABELS
] = { .minlen
= sizeof(struct md_labels
),
1076 .maxlen
= sizeof(struct md_labels
) },
1077 [OVS_CT_ATTR_HELPER
] = { .minlen
= 1,
1078 .maxlen
= NF_CT_HELPER_NAME_LEN
},
1079 #ifdef CONFIG_NF_NAT_NEEDED
1080 /* NAT length is checked when parsing the nested attributes. */
1081 [OVS_CT_ATTR_NAT
] = { .minlen
= 0, .maxlen
= INT_MAX
},
1085 static int parse_ct(const struct nlattr
*attr
, struct ovs_conntrack_info
*info
,
1086 const char **helper
, bool log
)
1091 nla_for_each_nested(a
, attr
, rem
) {
1092 int type
= nla_type(a
);
1093 int maxlen
= ovs_ct_attr_lens
[type
].maxlen
;
1094 int minlen
= ovs_ct_attr_lens
[type
].minlen
;
1096 if (type
> OVS_CT_ATTR_MAX
) {
1098 "Unknown conntrack attr (type=%d, max=%d)",
1099 type
, OVS_CT_ATTR_MAX
);
1102 if (nla_len(a
) < minlen
|| nla_len(a
) > maxlen
) {
1104 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1105 type
, nla_len(a
), maxlen
);
1110 case OVS_CT_ATTR_COMMIT
:
1111 info
->commit
= true;
1113 #ifdef CONFIG_NF_CONNTRACK_ZONES
1114 case OVS_CT_ATTR_ZONE
:
1115 info
->zone
.id
= nla_get_u16(a
);
1118 #ifdef CONFIG_NF_CONNTRACK_MARK
1119 case OVS_CT_ATTR_MARK
: {
1120 struct md_mark
*mark
= nla_data(a
);
1123 OVS_NLERR(log
, "ct_mark mask cannot be 0");
1130 #ifdef CONFIG_NF_CONNTRACK_LABELS
1131 case OVS_CT_ATTR_LABELS
: {
1132 struct md_labels
*labels
= nla_data(a
);
1134 if (!labels_nonzero(&labels
->mask
)) {
1135 OVS_NLERR(log
, "ct_labels mask cannot be 0");
1138 info
->labels
= *labels
;
1142 case OVS_CT_ATTR_HELPER
:
1143 *helper
= nla_data(a
);
1144 if (!memchr(*helper
, '\0', nla_len(a
))) {
1145 OVS_NLERR(log
, "Invalid conntrack helper");
1149 #ifdef CONFIG_NF_NAT_NEEDED
1150 case OVS_CT_ATTR_NAT
: {
1151 int err
= parse_nat(a
, info
, log
);
1159 OVS_NLERR(log
, "Unknown conntrack attr (%d)",
1165 #ifdef CONFIG_NF_CONNTRACK_MARK
1166 if (!info
->commit
&& info
->mark
.mask
) {
1168 "Setting conntrack mark requires 'commit' flag.");
1172 #ifdef CONFIG_NF_CONNTRACK_LABELS
1173 if (!info
->commit
&& labels_nonzero(&info
->labels
.mask
)) {
1175 "Setting conntrack labels requires 'commit' flag.");
1180 OVS_NLERR(log
, "Conntrack attr has %d unknown bytes", rem
);
1187 bool ovs_ct_verify(struct net
*net
, enum ovs_key_attr attr
)
1189 if (attr
== OVS_KEY_ATTR_CT_STATE
)
1191 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1192 attr
== OVS_KEY_ATTR_CT_ZONE
)
1194 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) &&
1195 attr
== OVS_KEY_ATTR_CT_MARK
)
1197 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1198 attr
== OVS_KEY_ATTR_CT_LABELS
) {
1199 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1201 return ovs_net
->xt_label
;
1207 int ovs_ct_copy_action(struct net
*net
, const struct nlattr
*attr
,
1208 const struct sw_flow_key
*key
,
1209 struct sw_flow_actions
**sfa
, bool log
)
1211 struct ovs_conntrack_info ct_info
;
1212 const char *helper
= NULL
;
1216 family
= key_to_nfproto(key
);
1217 if (family
== NFPROTO_UNSPEC
) {
1218 OVS_NLERR(log
, "ct family unspecified");
1222 memset(&ct_info
, 0, sizeof(ct_info
));
1223 ct_info
.family
= family
;
1225 nf_ct_zone_init(&ct_info
.zone
, NF_CT_DEFAULT_ZONE_ID
,
1226 NF_CT_DEFAULT_ZONE_DIR
, 0);
1228 err
= parse_ct(attr
, &ct_info
, &helper
, log
);
1232 /* Set up template for tracking connections in specific zones. */
1233 ct_info
.ct
= nf_ct_tmpl_alloc(net
, &ct_info
.zone
, GFP_KERNEL
);
1235 OVS_NLERR(log
, "Failed to allocate conntrack template");
1239 __set_bit(IPS_CONFIRMED_BIT
, &ct_info
.ct
->status
);
1240 nf_conntrack_get(&ct_info
.ct
->ct_general
);
1243 err
= ovs_ct_add_helper(&ct_info
, helper
, key
, log
);
1248 err
= ovs_nla_add_action(sfa
, OVS_ACTION_ATTR_CT
, &ct_info
,
1249 sizeof(ct_info
), log
);
1255 __ovs_ct_free_action(&ct_info
);
1259 #ifdef CONFIG_NF_NAT_NEEDED
1260 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info
*info
,
1261 struct sk_buff
*skb
)
1263 struct nlattr
*start
;
1265 start
= nla_nest_start(skb
, OVS_CT_ATTR_NAT
);
1269 if (info
->nat
& OVS_CT_SRC_NAT
) {
1270 if (nla_put_flag(skb
, OVS_NAT_ATTR_SRC
))
1272 } else if (info
->nat
& OVS_CT_DST_NAT
) {
1273 if (nla_put_flag(skb
, OVS_NAT_ATTR_DST
))
1279 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
) {
1280 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
1281 info
->family
== NFPROTO_IPV4
) {
1282 if (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1283 info
->range
.min_addr
.ip
) ||
1284 (info
->range
.max_addr
.ip
1285 != info
->range
.min_addr
.ip
&&
1286 (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1287 info
->range
.max_addr
.ip
))))
1289 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
1290 info
->family
== NFPROTO_IPV6
) {
1291 if (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1292 &info
->range
.min_addr
.in6
) ||
1293 (memcmp(&info
->range
.max_addr
.in6
,
1294 &info
->range
.min_addr
.in6
,
1295 sizeof(info
->range
.max_addr
.in6
)) &&
1296 (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1297 &info
->range
.max_addr
.in6
))))
1303 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1304 (nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MIN
,
1305 ntohs(info
->range
.min_proto
.all
)) ||
1306 (info
->range
.max_proto
.all
!= info
->range
.min_proto
.all
&&
1307 nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MAX
,
1308 ntohs(info
->range
.max_proto
.all
)))))
1311 if (info
->range
.flags
& NF_NAT_RANGE_PERSISTENT
&&
1312 nla_put_flag(skb
, OVS_NAT_ATTR_PERSISTENT
))
1314 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM
&&
1315 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_HASH
))
1317 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM_FULLY
&&
1318 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_RANDOM
))
1321 nla_nest_end(skb
, start
);
1327 int ovs_ct_action_to_attr(const struct ovs_conntrack_info
*ct_info
,
1328 struct sk_buff
*skb
)
1330 struct nlattr
*start
;
1332 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_CT
);
1336 if (ct_info
->commit
&& nla_put_flag(skb
, OVS_CT_ATTR_COMMIT
))
1338 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1339 nla_put_u16(skb
, OVS_CT_ATTR_ZONE
, ct_info
->zone
.id
))
1341 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) && ct_info
->mark
.mask
&&
1342 nla_put(skb
, OVS_CT_ATTR_MARK
, sizeof(ct_info
->mark
),
1345 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1346 labels_nonzero(&ct_info
->labels
.mask
) &&
1347 nla_put(skb
, OVS_CT_ATTR_LABELS
, sizeof(ct_info
->labels
),
1350 if (ct_info
->helper
) {
1351 if (nla_put_string(skb
, OVS_CT_ATTR_HELPER
,
1352 ct_info
->helper
->name
))
1355 #ifdef CONFIG_NF_NAT_NEEDED
1356 if (ct_info
->nat
&& !ovs_ct_nat_to_attr(ct_info
, skb
))
1359 nla_nest_end(skb
, start
);
1364 void ovs_ct_free_action(const struct nlattr
*a
)
1366 struct ovs_conntrack_info
*ct_info
= nla_data(a
);
1368 __ovs_ct_free_action(ct_info
);
1371 static void __ovs_ct_free_action(struct ovs_conntrack_info
*ct_info
)
1373 if (ct_info
->helper
)
1374 module_put(ct_info
->helper
->me
);
1376 nf_ct_put(ct_info
->ct
);
1379 void ovs_ct_init(struct net
*net
)
1381 unsigned int n_bits
= sizeof(struct ovs_key_ct_labels
) * BITS_PER_BYTE
;
1382 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1384 if (nf_connlabels_get(net
, n_bits
- 1)) {
1385 ovs_net
->xt_label
= false;
1386 OVS_NLERR(true, "Failed to set connlabel length");
1388 ovs_net
->xt_label
= true;
1392 void ovs_ct_exit(struct net
*net
)
1394 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1396 if (ovs_net
->xt_label
)
1397 nf_connlabels_put(net
);