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 */
71 struct md_labels labels
;
72 #ifdef CONFIG_NF_NAT_NEEDED
73 struct nf_nat_range range
; /* Only present for SRC NAT and DST NAT. */
77 static bool labels_nonzero(const struct ovs_key_ct_labels
*labels
);
79 static void __ovs_ct_free_action(struct ovs_conntrack_info
*ct_info
);
81 static u16
key_to_nfproto(const struct sw_flow_key
*key
)
83 switch (ntohs(key
->eth
.type
)) {
89 return NFPROTO_UNSPEC
;
93 /* Map SKB connection state into the values used by flow definition. */
94 static u8
ovs_ct_get_state(enum ip_conntrack_info ctinfo
)
96 u8 ct_state
= OVS_CS_F_TRACKED
;
99 case IP_CT_ESTABLISHED_REPLY
:
100 case IP_CT_RELATED_REPLY
:
101 ct_state
|= OVS_CS_F_REPLY_DIR
;
108 case IP_CT_ESTABLISHED
:
109 case IP_CT_ESTABLISHED_REPLY
:
110 ct_state
|= OVS_CS_F_ESTABLISHED
;
113 case IP_CT_RELATED_REPLY
:
114 ct_state
|= OVS_CS_F_RELATED
;
117 ct_state
|= OVS_CS_F_NEW
;
126 static u32
ovs_ct_get_mark(const struct nf_conn
*ct
)
128 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
129 return ct
? ct
->mark
: 0;
135 /* Guard against conntrack labels max size shrinking below 128 bits. */
136 #if NF_CT_LABELS_MAX_SIZE < 16
137 #error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes
140 static void ovs_ct_get_labels(const struct nf_conn
*ct
,
141 struct ovs_key_ct_labels
*labels
)
143 struct nf_conn_labels
*cl
= ct
? nf_ct_labels_find(ct
) : NULL
;
146 memcpy(labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
148 memset(labels
, 0, OVS_CT_LABELS_LEN
);
151 static void __ovs_ct_update_key_orig_tp(struct sw_flow_key
*key
,
152 const struct nf_conntrack_tuple
*orig
,
155 key
->ct_orig_proto
= orig
->dst
.protonum
;
156 if (orig
->dst
.protonum
== icmp_proto
) {
157 key
->ct
.orig_tp
.src
= htons(orig
->dst
.u
.icmp
.type
);
158 key
->ct
.orig_tp
.dst
= htons(orig
->dst
.u
.icmp
.code
);
160 key
->ct
.orig_tp
.src
= orig
->src
.u
.all
;
161 key
->ct
.orig_tp
.dst
= orig
->dst
.u
.all
;
165 static void __ovs_ct_update_key(struct sw_flow_key
*key
, u8 state
,
166 const struct nf_conntrack_zone
*zone
,
167 const struct nf_conn
*ct
)
169 key
->ct_state
= state
;
170 key
->ct_zone
= zone
->id
;
171 key
->ct
.mark
= ovs_ct_get_mark(ct
);
172 ovs_ct_get_labels(ct
, &key
->ct
.labels
);
175 const struct nf_conntrack_tuple
*orig
;
177 /* Use the master if we have one. */
180 orig
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
182 /* IP version must match with the master connection. */
183 if (key
->eth
.type
== htons(ETH_P_IP
) &&
184 nf_ct_l3num(ct
) == NFPROTO_IPV4
) {
185 key
->ipv4
.ct_orig
.src
= orig
->src
.u3
.ip
;
186 key
->ipv4
.ct_orig
.dst
= orig
->dst
.u3
.ip
;
187 __ovs_ct_update_key_orig_tp(key
, orig
, IPPROTO_ICMP
);
189 } else if (key
->eth
.type
== htons(ETH_P_IPV6
) &&
190 !sw_flow_key_is_nd(key
) &&
191 nf_ct_l3num(ct
) == NFPROTO_IPV6
) {
192 key
->ipv6
.ct_orig
.src
= orig
->src
.u3
.in6
;
193 key
->ipv6
.ct_orig
.dst
= orig
->dst
.u3
.in6
;
194 __ovs_ct_update_key_orig_tp(key
, orig
, NEXTHDR_ICMP
);
198 /* Clear 'ct_orig_proto' to mark the non-existence of conntrack
199 * original direction key fields.
201 key
->ct_orig_proto
= 0;
204 /* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has
205 * previously sent the packet to conntrack via the ct action. If
206 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
207 * initialized from the connection status.
209 static void ovs_ct_update_key(const struct sk_buff
*skb
,
210 const struct ovs_conntrack_info
*info
,
211 struct sw_flow_key
*key
, bool post_ct
,
214 const struct nf_conntrack_zone
*zone
= &nf_ct_zone_dflt
;
215 enum ip_conntrack_info ctinfo
;
219 ct
= nf_ct_get(skb
, &ctinfo
);
221 state
= ovs_ct_get_state(ctinfo
);
222 /* All unconfirmed entries are NEW connections. */
223 if (!nf_ct_is_confirmed(ct
))
224 state
|= OVS_CS_F_NEW
;
225 /* OVS persists the related flag for the duration of the
229 state
|= OVS_CS_F_RELATED
;
230 if (keep_nat_flags
) {
231 state
|= key
->ct_state
& OVS_CS_F_NAT_MASK
;
233 if (ct
->status
& IPS_SRC_NAT
)
234 state
|= OVS_CS_F_SRC_NAT
;
235 if (ct
->status
& IPS_DST_NAT
)
236 state
|= OVS_CS_F_DST_NAT
;
238 zone
= nf_ct_zone(ct
);
239 } else if (post_ct
) {
240 state
= OVS_CS_F_TRACKED
| OVS_CS_F_INVALID
;
244 __ovs_ct_update_key(key
, state
, zone
, ct
);
247 /* This is called to initialize CT key fields possibly coming in from the local
250 void ovs_ct_fill_key(const struct sk_buff
*skb
, struct sw_flow_key
*key
)
252 ovs_ct_update_key(skb
, NULL
, key
, false, false);
255 #define IN6_ADDR_INITIALIZER(ADDR) \
256 { (ADDR).s6_addr32[0], (ADDR).s6_addr32[1], \
257 (ADDR).s6_addr32[2], (ADDR).s6_addr32[3] }
259 int ovs_ct_put_key(const struct sw_flow_key
*swkey
,
260 const struct sw_flow_key
*output
, struct sk_buff
*skb
)
262 if (nla_put_u32(skb
, OVS_KEY_ATTR_CT_STATE
, output
->ct_state
))
265 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
266 nla_put_u16(skb
, OVS_KEY_ATTR_CT_ZONE
, output
->ct_zone
))
269 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) &&
270 nla_put_u32(skb
, OVS_KEY_ATTR_CT_MARK
, output
->ct
.mark
))
273 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
274 nla_put(skb
, OVS_KEY_ATTR_CT_LABELS
, sizeof(output
->ct
.labels
),
278 if (swkey
->ct_orig_proto
) {
279 if (swkey
->eth
.type
== htons(ETH_P_IP
)) {
280 struct ovs_key_ct_tuple_ipv4 orig
= {
281 output
->ipv4
.ct_orig
.src
,
282 output
->ipv4
.ct_orig
.dst
,
283 output
->ct
.orig_tp
.src
,
284 output
->ct
.orig_tp
.dst
,
285 output
->ct_orig_proto
,
287 if (nla_put(skb
, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
,
288 sizeof(orig
), &orig
))
290 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
)) {
291 struct ovs_key_ct_tuple_ipv6 orig
= {
292 IN6_ADDR_INITIALIZER(output
->ipv6
.ct_orig
.src
),
293 IN6_ADDR_INITIALIZER(output
->ipv6
.ct_orig
.dst
),
294 output
->ct
.orig_tp
.src
,
295 output
->ct
.orig_tp
.dst
,
296 output
->ct_orig_proto
,
298 if (nla_put(skb
, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
,
299 sizeof(orig
), &orig
))
307 static int ovs_ct_set_mark(struct nf_conn
*ct
, struct sw_flow_key
*key
,
308 u32 ct_mark
, u32 mask
)
310 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
313 new_mark
= ct_mark
| (ct
->mark
& ~(mask
));
314 if (ct
->mark
!= new_mark
) {
316 if (nf_ct_is_confirmed(ct
))
317 nf_conntrack_event_cache(IPCT_MARK
, ct
);
318 key
->ct
.mark
= new_mark
;
327 static struct nf_conn_labels
*ovs_ct_get_conn_labels(struct nf_conn
*ct
)
329 struct nf_conn_labels
*cl
;
331 cl
= nf_ct_labels_find(ct
);
333 nf_ct_labels_ext_add(ct
);
334 cl
= nf_ct_labels_find(ct
);
340 /* Initialize labels for a new, yet to be committed conntrack entry. Note that
341 * since the new connection is not yet confirmed, and thus no-one else has
342 * access to it's labels, we simply write them over.
344 static int ovs_ct_init_labels(struct nf_conn
*ct
, struct sw_flow_key
*key
,
345 const struct ovs_key_ct_labels
*labels
,
346 const struct ovs_key_ct_labels
*mask
)
348 struct nf_conn_labels
*cl
, *master_cl
;
349 bool have_mask
= labels_nonzero(mask
);
351 /* Inherit master's labels to the related connection? */
352 master_cl
= ct
->master
? nf_ct_labels_find(ct
->master
) : NULL
;
354 if (!master_cl
&& !have_mask
)
355 return 0; /* Nothing to do. */
357 cl
= ovs_ct_get_conn_labels(ct
);
361 /* Inherit the master's labels, if any. */
366 u32
*dst
= (u32
*)cl
->bits
;
369 for (i
= 0; i
< OVS_CT_LABELS_LEN_32
; i
++)
370 dst
[i
] = (dst
[i
] & ~mask
->ct_labels_32
[i
]) |
371 (labels
->ct_labels_32
[i
]
372 & mask
->ct_labels_32
[i
]);
375 /* Labels are included in the IPCTNL_MSG_CT_NEW event only if the
376 * IPCT_LABEL bit it set in the event cache.
378 nf_conntrack_event_cache(IPCT_LABEL
, ct
);
380 memcpy(&key
->ct
.labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
385 static int ovs_ct_set_labels(struct nf_conn
*ct
, struct sw_flow_key
*key
,
386 const struct ovs_key_ct_labels
*labels
,
387 const struct ovs_key_ct_labels
*mask
)
389 struct nf_conn_labels
*cl
;
392 cl
= ovs_ct_get_conn_labels(ct
);
396 err
= nf_connlabels_replace(ct
, labels
->ct_labels_32
,
398 OVS_CT_LABELS_LEN_32
);
402 memcpy(&key
->ct
.labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
407 /* 'skb' should already be pulled to nh_ofs. */
408 static int ovs_ct_helper(struct sk_buff
*skb
, u16 proto
)
410 const struct nf_conntrack_helper
*helper
;
411 const struct nf_conn_help
*help
;
412 enum ip_conntrack_info ctinfo
;
413 unsigned int protoff
;
417 ct
= nf_ct_get(skb
, &ctinfo
);
418 if (!ct
|| ctinfo
== IP_CT_RELATED_REPLY
)
421 help
= nfct_help(ct
);
425 helper
= rcu_dereference(help
->helper
);
431 protoff
= ip_hdrlen(skb
);
434 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
438 ofs
= ipv6_skip_exthdr(skb
, sizeof(struct ipv6hdr
), &nexthdr
,
440 if (ofs
< 0 || (frag_off
& htons(~0x7)) != 0) {
441 pr_debug("proto header not found\n");
448 WARN_ONCE(1, "helper invoked on non-IP family!");
452 err
= helper
->help(skb
, protoff
, ct
, ctinfo
);
453 if (err
!= NF_ACCEPT
)
456 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
457 * FTP with NAT) adusting the TCP payload size when mangling IP
458 * addresses and/or port numbers in the text-based control connection.
460 if (test_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
) &&
461 !nf_ct_seq_adjust(skb
, ct
, ctinfo
, protoff
))
466 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
467 * value if 'skb' is freed.
469 static int handle_fragments(struct net
*net
, struct sw_flow_key
*key
,
470 u16 zone
, struct sk_buff
*skb
)
472 struct ovs_skb_cb ovs_cb
= *OVS_CB(skb
);
475 if (key
->eth
.type
== htons(ETH_P_IP
)) {
476 enum ip_defrag_users user
= IP_DEFRAG_CONNTRACK_IN
+ zone
;
478 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
479 err
= ip_defrag(net
, skb
, user
);
483 ovs_cb
.mru
= IPCB(skb
)->frag_max_size
;
484 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
485 } else if (key
->eth
.type
== htons(ETH_P_IPV6
)) {
486 enum ip6_defrag_users user
= IP6_DEFRAG_CONNTRACK_IN
+ zone
;
488 memset(IP6CB(skb
), 0, sizeof(struct inet6_skb_parm
));
489 err
= nf_ct_frag6_gather(net
, skb
, user
);
491 if (err
!= -EINPROGRESS
)
496 key
->ip
.proto
= ipv6_hdr(skb
)->nexthdr
;
497 ovs_cb
.mru
= IP6CB(skb
)->frag_max_size
;
501 return -EPFNOSUPPORT
;
504 key
->ip
.frag
= OVS_FRAG_TYPE_NONE
;
507 *OVS_CB(skb
) = ovs_cb
;
512 static struct nf_conntrack_expect
*
513 ovs_ct_expect_find(struct net
*net
, const struct nf_conntrack_zone
*zone
,
514 u16 proto
, const struct sk_buff
*skb
)
516 struct nf_conntrack_tuple tuple
;
518 if (!nf_ct_get_tuplepr(skb
, skb_network_offset(skb
), proto
, net
, &tuple
))
520 return __nf_ct_expect_find(net
, zone
, &tuple
);
523 /* This replicates logic from nf_conntrack_core.c that is not exported. */
524 static enum ip_conntrack_info
525 ovs_ct_get_info(const struct nf_conntrack_tuple_hash
*h
)
527 const struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(h
);
529 if (NF_CT_DIRECTION(h
) == IP_CT_DIR_REPLY
)
530 return IP_CT_ESTABLISHED_REPLY
;
531 /* Once we've had two way comms, always ESTABLISHED. */
532 if (test_bit(IPS_SEEN_REPLY_BIT
, &ct
->status
))
533 return IP_CT_ESTABLISHED
;
534 if (test_bit(IPS_EXPECTED_BIT
, &ct
->status
))
535 return IP_CT_RELATED
;
539 /* Find an existing connection which this packet belongs to without
540 * re-attributing statistics or modifying the connection state. This allows an
541 * skb->_nfct lost due to an upcall to be recovered during actions execution.
543 * Must be called with rcu_read_lock.
545 * On success, populates skb->_nfct and returns the connection. Returns NULL
546 * if there is no existing entry.
548 static struct nf_conn
*
549 ovs_ct_find_existing(struct net
*net
, const struct nf_conntrack_zone
*zone
,
550 u8 l3num
, struct sk_buff
*skb
, bool natted
)
552 struct nf_conntrack_l3proto
*l3proto
;
553 struct nf_conntrack_l4proto
*l4proto
;
554 struct nf_conntrack_tuple tuple
;
555 struct nf_conntrack_tuple_hash
*h
;
557 unsigned int dataoff
;
560 l3proto
= __nf_ct_l3proto_find(l3num
);
561 if (l3proto
->get_l4proto(skb
, skb_network_offset(skb
), &dataoff
,
563 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
566 l4proto
= __nf_ct_l4proto_find(l3num
, protonum
);
567 if (!nf_ct_get_tuple(skb
, skb_network_offset(skb
), dataoff
, l3num
,
568 protonum
, net
, &tuple
, l3proto
, l4proto
)) {
569 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
573 /* Must invert the tuple if skb has been transformed by NAT. */
575 struct nf_conntrack_tuple inverse
;
577 if (!nf_ct_invert_tuple(&inverse
, &tuple
, l3proto
, l4proto
)) {
578 pr_debug("ovs_ct_find_existing: Inversion failed!\n");
584 /* look for tuple match */
585 h
= nf_conntrack_find_get(net
, zone
, &tuple
);
587 return NULL
; /* Not found. */
589 ct
= nf_ct_tuplehash_to_ctrack(h
);
591 /* Inverted packet tuple matches the reverse direction conntrack tuple,
592 * select the other tuplehash to get the right 'ctinfo' bits for this
596 h
= &ct
->tuplehash
[!h
->tuple
.dst
.dir
];
598 nf_ct_set(skb
, ct
, ovs_ct_get_info(h
));
602 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
603 static bool skb_nfct_cached(struct net
*net
,
604 const struct sw_flow_key
*key
,
605 const struct ovs_conntrack_info
*info
,
608 enum ip_conntrack_info ctinfo
;
611 ct
= nf_ct_get(skb
, &ctinfo
);
612 /* If no ct, check if we have evidence that an existing conntrack entry
613 * might be found for this skb. This happens when we lose a skb->_nfct
614 * due to an upcall. If the connection was not confirmed, it is not
615 * cached and needs to be run through conntrack again.
617 if (!ct
&& key
->ct_state
& OVS_CS_F_TRACKED
&&
618 !(key
->ct_state
& OVS_CS_F_INVALID
) &&
619 key
->ct_zone
== info
->zone
.id
) {
620 ct
= ovs_ct_find_existing(net
, &info
->zone
, info
->family
, skb
,
622 & OVS_CS_F_NAT_MASK
));
624 nf_ct_get(skb
, &ctinfo
);
628 if (!net_eq(net
, read_pnet(&ct
->ct_net
)))
630 if (!nf_ct_zone_equal_any(info
->ct
, nf_ct_zone(ct
)))
633 struct nf_conn_help
*help
;
635 help
= nf_ct_ext_find(ct
, NF_CT_EXT_HELPER
);
636 if (help
&& rcu_access_pointer(help
->helper
) != info
->helper
)
639 /* Force conntrack entry direction to the current packet? */
640 if (info
->force
&& CTINFO2DIR(ctinfo
) != IP_CT_DIR_ORIGINAL
) {
641 /* Delete the conntrack entry if confirmed, else just release
644 if (nf_ct_is_confirmed(ct
))
645 nf_ct_delete(ct
, 0, 0);
647 nf_conntrack_put(&ct
->ct_general
);
648 nf_ct_set(skb
, NULL
, 0);
655 #ifdef CONFIG_NF_NAT_NEEDED
656 /* Modelled after nf_nat_ipv[46]_fn().
657 * range is only used for new, uninitialized NAT state.
658 * Returns either NF_ACCEPT or NF_DROP.
660 static int ovs_ct_nat_execute(struct sk_buff
*skb
, struct nf_conn
*ct
,
661 enum ip_conntrack_info ctinfo
,
662 const struct nf_nat_range
*range
,
663 enum nf_nat_manip_type maniptype
)
665 int hooknum
, nh_off
, err
= NF_ACCEPT
;
667 nh_off
= skb_network_offset(skb
);
668 skb_pull_rcsum(skb
, nh_off
);
670 /* See HOOK2MANIP(). */
671 if (maniptype
== NF_NAT_MANIP_SRC
)
672 hooknum
= NF_INET_LOCAL_IN
; /* Source NAT */
674 hooknum
= NF_INET_LOCAL_OUT
; /* Destination NAT */
678 case IP_CT_RELATED_REPLY
:
679 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
680 skb
->protocol
== htons(ETH_P_IP
) &&
681 ip_hdr(skb
)->protocol
== IPPROTO_ICMP
) {
682 if (!nf_nat_icmp_reply_translation(skb
, ct
, ctinfo
,
686 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
687 skb
->protocol
== htons(ETH_P_IPV6
)) {
689 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
690 int hdrlen
= ipv6_skip_exthdr(skb
,
691 sizeof(struct ipv6hdr
),
692 &nexthdr
, &frag_off
);
694 if (hdrlen
>= 0 && nexthdr
== IPPROTO_ICMPV6
) {
695 if (!nf_nat_icmpv6_reply_translation(skb
, ct
,
703 /* Non-ICMP, fall thru to initialize if needed. */
705 /* Seen it before? This can happen for loopback, retrans,
708 if (!nf_nat_initialized(ct
, maniptype
)) {
709 /* Initialize according to the NAT action. */
710 err
= (range
&& range
->flags
& NF_NAT_RANGE_MAP_IPS
)
711 /* Action is set up to establish a new
714 ? nf_nat_setup_info(ct
, range
, maniptype
)
715 : nf_nat_alloc_null_binding(ct
, hooknum
);
716 if (err
!= NF_ACCEPT
)
721 case IP_CT_ESTABLISHED
:
722 case IP_CT_ESTABLISHED_REPLY
:
730 err
= nf_nat_packet(ct
, ctinfo
, hooknum
, skb
);
732 skb_push(skb
, nh_off
);
733 skb_postpush_rcsum(skb
, skb
->data
, nh_off
);
738 static void ovs_nat_update_key(struct sw_flow_key
*key
,
739 const struct sk_buff
*skb
,
740 enum nf_nat_manip_type maniptype
)
742 if (maniptype
== NF_NAT_MANIP_SRC
) {
745 key
->ct_state
|= OVS_CS_F_SRC_NAT
;
746 if (key
->eth
.type
== htons(ETH_P_IP
))
747 key
->ipv4
.addr
.src
= ip_hdr(skb
)->saddr
;
748 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
749 memcpy(&key
->ipv6
.addr
.src
, &ipv6_hdr(skb
)->saddr
,
750 sizeof(key
->ipv6
.addr
.src
));
754 if (key
->ip
.proto
== IPPROTO_UDP
)
755 src
= udp_hdr(skb
)->source
;
756 else if (key
->ip
.proto
== IPPROTO_TCP
)
757 src
= tcp_hdr(skb
)->source
;
758 else if (key
->ip
.proto
== IPPROTO_SCTP
)
759 src
= sctp_hdr(skb
)->source
;
767 key
->ct_state
|= OVS_CS_F_DST_NAT
;
768 if (key
->eth
.type
== htons(ETH_P_IP
))
769 key
->ipv4
.addr
.dst
= ip_hdr(skb
)->daddr
;
770 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
771 memcpy(&key
->ipv6
.addr
.dst
, &ipv6_hdr(skb
)->daddr
,
772 sizeof(key
->ipv6
.addr
.dst
));
776 if (key
->ip
.proto
== IPPROTO_UDP
)
777 dst
= udp_hdr(skb
)->dest
;
778 else if (key
->ip
.proto
== IPPROTO_TCP
)
779 dst
= tcp_hdr(skb
)->dest
;
780 else if (key
->ip
.proto
== IPPROTO_SCTP
)
781 dst
= sctp_hdr(skb
)->dest
;
789 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
790 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
791 const struct ovs_conntrack_info
*info
,
792 struct sk_buff
*skb
, struct nf_conn
*ct
,
793 enum ip_conntrack_info ctinfo
)
795 enum nf_nat_manip_type maniptype
;
798 if (nf_ct_is_untracked(ct
)) {
799 /* A NAT action may only be performed on tracked packets. */
803 /* Add NAT extension if not confirmed yet. */
804 if (!nf_ct_is_confirmed(ct
) && !nf_ct_nat_ext_add(ct
))
805 return NF_ACCEPT
; /* Can't NAT. */
807 /* Determine NAT type.
808 * Check if the NAT type can be deduced from the tracked connection.
809 * Make sure new expected connections (IP_CT_RELATED) are NATted only
812 if (info
->nat
& OVS_CT_NAT
&& ctinfo
!= IP_CT_NEW
&&
813 ct
->status
& IPS_NAT_MASK
&&
814 (ctinfo
!= IP_CT_RELATED
|| info
->commit
)) {
815 /* NAT an established or related connection like before. */
816 if (CTINFO2DIR(ctinfo
) == IP_CT_DIR_REPLY
)
817 /* This is the REPLY direction for a connection
818 * for which NAT was applied in the forward
819 * direction. Do the reverse NAT.
821 maniptype
= ct
->status
& IPS_SRC_NAT
822 ? NF_NAT_MANIP_DST
: NF_NAT_MANIP_SRC
;
824 maniptype
= ct
->status
& IPS_SRC_NAT
825 ? NF_NAT_MANIP_SRC
: NF_NAT_MANIP_DST
;
826 } else if (info
->nat
& OVS_CT_SRC_NAT
) {
827 maniptype
= NF_NAT_MANIP_SRC
;
828 } else if (info
->nat
& OVS_CT_DST_NAT
) {
829 maniptype
= NF_NAT_MANIP_DST
;
831 return NF_ACCEPT
; /* Connection is not NATed. */
833 err
= ovs_ct_nat_execute(skb
, ct
, ctinfo
, &info
->range
, maniptype
);
835 /* Mark NAT done if successful and update the flow key. */
836 if (err
== NF_ACCEPT
)
837 ovs_nat_update_key(key
, skb
, maniptype
);
841 #else /* !CONFIG_NF_NAT_NEEDED */
842 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
843 const struct ovs_conntrack_info
*info
,
844 struct sk_buff
*skb
, struct nf_conn
*ct
,
845 enum ip_conntrack_info ctinfo
)
851 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
852 * not done already. Update key with new CT state after passing the packet
854 * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be
855 * set to NULL and 0 will be returned.
857 static int __ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
858 const struct ovs_conntrack_info
*info
,
861 /* If we are recirculating packets to match on conntrack fields and
862 * committing with a separate conntrack action, then we don't need to
863 * actually run the packet through conntrack twice unless it's for a
866 bool cached
= skb_nfct_cached(net
, key
, info
, skb
);
867 enum ip_conntrack_info ctinfo
;
871 struct nf_conn
*tmpl
= info
->ct
;
874 /* Associate skb with specified zone. */
877 nf_conntrack_put(skb_nfct(skb
));
878 nf_conntrack_get(&tmpl
->ct_general
);
879 nf_ct_set(skb
, tmpl
, IP_CT_NEW
);
882 err
= nf_conntrack_in(net
, info
->family
,
883 NF_INET_PRE_ROUTING
, skb
);
884 if (err
!= NF_ACCEPT
)
887 /* Clear CT state NAT flags to mark that we have not yet done
888 * NAT after the nf_conntrack_in() call. We can actually clear
889 * the whole state, as it will be re-initialized below.
893 /* Update the key, but keep the NAT flags. */
894 ovs_ct_update_key(skb
, info
, key
, true, true);
897 ct
= nf_ct_get(skb
, &ctinfo
);
899 /* Packets starting a new connection must be NATted before the
900 * helper, so that the helper knows about the NAT. We enforce
901 * this by delaying both NAT and helper calls for unconfirmed
902 * connections until the committing CT action. For later
903 * packets NAT and Helper may be called in either order.
905 * NAT will be done only if the CT action has NAT, and only
906 * once per packet (per zone), as guarded by the NAT bits in
909 if (info
->nat
&& !(key
->ct_state
& OVS_CS_F_NAT_MASK
) &&
910 (nf_ct_is_confirmed(ct
) || info
->commit
) &&
911 ovs_ct_nat(net
, key
, info
, skb
, ct
, ctinfo
) != NF_ACCEPT
) {
915 /* Userspace may decide to perform a ct lookup without a helper
916 * specified followed by a (recirculate and) commit with one.
917 * Therefore, for unconfirmed connections which we will commit,
918 * we need to attach the helper here.
920 if (!nf_ct_is_confirmed(ct
) && info
->commit
&&
921 info
->helper
&& !nfct_help(ct
)) {
922 int err
= __nf_ct_try_assign_helper(ct
, info
->ct
,
928 /* Call the helper only if:
929 * - nf_conntrack_in() was executed above ("!cached") for a
930 * confirmed connection, or
931 * - When committing an unconfirmed connection.
933 if ((nf_ct_is_confirmed(ct
) ? !cached
: info
->commit
) &&
934 ovs_ct_helper(skb
, info
->family
) != NF_ACCEPT
) {
942 /* Lookup connection and read fields into key. */
943 static int ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
944 const struct ovs_conntrack_info
*info
,
947 struct nf_conntrack_expect
*exp
;
949 /* If we pass an expected packet through nf_conntrack_in() the
950 * expectation is typically removed, but the packet could still be
951 * lost in upcall processing. To prevent this from happening we
952 * perform an explicit expectation lookup. Expected connections are
953 * always new, and will be passed through conntrack only when they are
954 * committed, as it is OK to remove the expectation at that time.
956 exp
= ovs_ct_expect_find(net
, &info
->zone
, info
->family
, skb
);
960 /* NOTE: New connections are NATted and Helped only when
961 * committed, so we are not calling into NAT here.
963 state
= OVS_CS_F_TRACKED
| OVS_CS_F_NEW
| OVS_CS_F_RELATED
;
964 __ovs_ct_update_key(key
, state
, &info
->zone
, exp
->master
);
969 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
973 ct
= (struct nf_conn
*)skb_nfct(skb
);
975 nf_ct_deliver_cached_events(ct
);
981 static bool labels_nonzero(const struct ovs_key_ct_labels
*labels
)
985 for (i
= 0; i
< OVS_CT_LABELS_LEN_32
; i
++)
986 if (labels
->ct_labels_32
[i
])
992 /* Lookup connection and confirm if unconfirmed. */
993 static int ovs_ct_commit(struct net
*net
, struct sw_flow_key
*key
,
994 const struct ovs_conntrack_info
*info
,
997 enum ip_conntrack_info ctinfo
;
1001 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
1005 /* The connection could be invalid, in which case this is a no-op.*/
1006 ct
= nf_ct_get(skb
, &ctinfo
);
1010 /* Apply changes before confirming the connection so that the initial
1011 * conntrack NEW netlink event carries the values given in the CT
1014 if (info
->mark
.mask
) {
1015 err
= ovs_ct_set_mark(ct
, key
, info
->mark
.value
,
1020 if (!nf_ct_is_confirmed(ct
)) {
1021 err
= ovs_ct_init_labels(ct
, key
, &info
->labels
.value
,
1022 &info
->labels
.mask
);
1025 } else if (labels_nonzero(&info
->labels
.mask
)) {
1026 err
= ovs_ct_set_labels(ct
, key
, &info
->labels
.value
,
1027 &info
->labels
.mask
);
1031 /* This will take care of sending queued events even if the connection
1032 * is already confirmed.
1034 if (nf_conntrack_confirm(skb
) != NF_ACCEPT
)
1040 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
1041 * value if 'skb' is freed.
1043 int ovs_ct_execute(struct net
*net
, struct sk_buff
*skb
,
1044 struct sw_flow_key
*key
,
1045 const struct ovs_conntrack_info
*info
)
1050 /* The conntrack module expects to be working at L3. */
1051 nh_ofs
= skb_network_offset(skb
);
1052 skb_pull_rcsum(skb
, nh_ofs
);
1054 if (key
->ip
.frag
!= OVS_FRAG_TYPE_NONE
) {
1055 err
= handle_fragments(net
, key
, info
->zone
.id
, skb
);
1061 err
= ovs_ct_commit(net
, key
, info
, skb
);
1063 err
= ovs_ct_lookup(net
, key
, info
, skb
);
1065 skb_push(skb
, nh_ofs
);
1066 skb_postpush_rcsum(skb
, skb
->data
, nh_ofs
);
1072 static int ovs_ct_add_helper(struct ovs_conntrack_info
*info
, const char *name
,
1073 const struct sw_flow_key
*key
, bool log
)
1075 struct nf_conntrack_helper
*helper
;
1076 struct nf_conn_help
*help
;
1078 helper
= nf_conntrack_helper_try_module_get(name
, info
->family
,
1081 OVS_NLERR(log
, "Unknown helper \"%s\"", name
);
1085 help
= nf_ct_helper_ext_add(info
->ct
, helper
, GFP_KERNEL
);
1087 module_put(helper
->me
);
1091 rcu_assign_pointer(help
->helper
, helper
);
1092 info
->helper
= helper
;
1096 #ifdef CONFIG_NF_NAT_NEEDED
1097 static int parse_nat(const struct nlattr
*attr
,
1098 struct ovs_conntrack_info
*info
, bool log
)
1102 bool have_ip_max
= false;
1103 bool have_proto_max
= false;
1104 bool ip_vers
= (info
->family
== NFPROTO_IPV6
);
1106 nla_for_each_nested(a
, attr
, rem
) {
1107 static const int ovs_nat_attr_lens
[OVS_NAT_ATTR_MAX
+ 1][2] = {
1108 [OVS_NAT_ATTR_SRC
] = {0, 0},
1109 [OVS_NAT_ATTR_DST
] = {0, 0},
1110 [OVS_NAT_ATTR_IP_MIN
] = {sizeof(struct in_addr
),
1111 sizeof(struct in6_addr
)},
1112 [OVS_NAT_ATTR_IP_MAX
] = {sizeof(struct in_addr
),
1113 sizeof(struct in6_addr
)},
1114 [OVS_NAT_ATTR_PROTO_MIN
] = {sizeof(u16
), sizeof(u16
)},
1115 [OVS_NAT_ATTR_PROTO_MAX
] = {sizeof(u16
), sizeof(u16
)},
1116 [OVS_NAT_ATTR_PERSISTENT
] = {0, 0},
1117 [OVS_NAT_ATTR_PROTO_HASH
] = {0, 0},
1118 [OVS_NAT_ATTR_PROTO_RANDOM
] = {0, 0},
1120 int type
= nla_type(a
);
1122 if (type
> OVS_NAT_ATTR_MAX
) {
1124 "Unknown NAT attribute (type=%d, max=%d).\n",
1125 type
, OVS_NAT_ATTR_MAX
);
1129 if (nla_len(a
) != ovs_nat_attr_lens
[type
][ip_vers
]) {
1131 "NAT attribute type %d has unexpected length (%d != %d).\n",
1133 ovs_nat_attr_lens
[type
][ip_vers
]);
1138 case OVS_NAT_ATTR_SRC
:
1139 case OVS_NAT_ATTR_DST
:
1142 "Only one type of NAT may be specified.\n"
1146 info
->nat
|= OVS_CT_NAT
;
1147 info
->nat
|= ((type
== OVS_NAT_ATTR_SRC
)
1148 ? OVS_CT_SRC_NAT
: OVS_CT_DST_NAT
);
1151 case OVS_NAT_ATTR_IP_MIN
:
1152 nla_memcpy(&info
->range
.min_addr
, a
,
1153 sizeof(info
->range
.min_addr
));
1154 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1157 case OVS_NAT_ATTR_IP_MAX
:
1159 nla_memcpy(&info
->range
.max_addr
, a
,
1160 sizeof(info
->range
.max_addr
));
1161 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1164 case OVS_NAT_ATTR_PROTO_MIN
:
1165 info
->range
.min_proto
.all
= htons(nla_get_u16(a
));
1166 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1169 case OVS_NAT_ATTR_PROTO_MAX
:
1170 have_proto_max
= true;
1171 info
->range
.max_proto
.all
= htons(nla_get_u16(a
));
1172 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1175 case OVS_NAT_ATTR_PERSISTENT
:
1176 info
->range
.flags
|= NF_NAT_RANGE_PERSISTENT
;
1179 case OVS_NAT_ATTR_PROTO_HASH
:
1180 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM
;
1183 case OVS_NAT_ATTR_PROTO_RANDOM
:
1184 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM_FULLY
;
1188 OVS_NLERR(log
, "Unknown nat attribute (%d).\n", type
);
1194 OVS_NLERR(log
, "NAT attribute has %d unknown bytes.\n", rem
);
1198 /* Do not allow flags if no type is given. */
1199 if (info
->range
.flags
) {
1201 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1205 info
->nat
= OVS_CT_NAT
; /* NAT existing connections. */
1206 } else if (!info
->commit
) {
1208 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1212 /* Allow missing IP_MAX. */
1213 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
&& !have_ip_max
) {
1214 memcpy(&info
->range
.max_addr
, &info
->range
.min_addr
,
1215 sizeof(info
->range
.max_addr
));
1217 /* Allow missing PROTO_MAX. */
1218 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1220 info
->range
.max_proto
.all
= info
->range
.min_proto
.all
;
1226 static const struct ovs_ct_len_tbl ovs_ct_attr_lens
[OVS_CT_ATTR_MAX
+ 1] = {
1227 [OVS_CT_ATTR_COMMIT
] = { .minlen
= 0, .maxlen
= 0 },
1228 [OVS_CT_ATTR_FORCE_COMMIT
] = { .minlen
= 0, .maxlen
= 0 },
1229 [OVS_CT_ATTR_ZONE
] = { .minlen
= sizeof(u16
),
1230 .maxlen
= sizeof(u16
) },
1231 [OVS_CT_ATTR_MARK
] = { .minlen
= sizeof(struct md_mark
),
1232 .maxlen
= sizeof(struct md_mark
) },
1233 [OVS_CT_ATTR_LABELS
] = { .minlen
= sizeof(struct md_labels
),
1234 .maxlen
= sizeof(struct md_labels
) },
1235 [OVS_CT_ATTR_HELPER
] = { .minlen
= 1,
1236 .maxlen
= NF_CT_HELPER_NAME_LEN
},
1237 #ifdef CONFIG_NF_NAT_NEEDED
1238 /* NAT length is checked when parsing the nested attributes. */
1239 [OVS_CT_ATTR_NAT
] = { .minlen
= 0, .maxlen
= INT_MAX
},
1243 static int parse_ct(const struct nlattr
*attr
, struct ovs_conntrack_info
*info
,
1244 const char **helper
, bool log
)
1249 nla_for_each_nested(a
, attr
, rem
) {
1250 int type
= nla_type(a
);
1251 int maxlen
= ovs_ct_attr_lens
[type
].maxlen
;
1252 int minlen
= ovs_ct_attr_lens
[type
].minlen
;
1254 if (type
> OVS_CT_ATTR_MAX
) {
1256 "Unknown conntrack attr (type=%d, max=%d)",
1257 type
, OVS_CT_ATTR_MAX
);
1260 if (nla_len(a
) < minlen
|| nla_len(a
) > maxlen
) {
1262 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1263 type
, nla_len(a
), maxlen
);
1268 case OVS_CT_ATTR_FORCE_COMMIT
:
1271 case OVS_CT_ATTR_COMMIT
:
1272 info
->commit
= true;
1274 #ifdef CONFIG_NF_CONNTRACK_ZONES
1275 case OVS_CT_ATTR_ZONE
:
1276 info
->zone
.id
= nla_get_u16(a
);
1279 #ifdef CONFIG_NF_CONNTRACK_MARK
1280 case OVS_CT_ATTR_MARK
: {
1281 struct md_mark
*mark
= nla_data(a
);
1284 OVS_NLERR(log
, "ct_mark mask cannot be 0");
1291 #ifdef CONFIG_NF_CONNTRACK_LABELS
1292 case OVS_CT_ATTR_LABELS
: {
1293 struct md_labels
*labels
= nla_data(a
);
1295 if (!labels_nonzero(&labels
->mask
)) {
1296 OVS_NLERR(log
, "ct_labels mask cannot be 0");
1299 info
->labels
= *labels
;
1303 case OVS_CT_ATTR_HELPER
:
1304 *helper
= nla_data(a
);
1305 if (!memchr(*helper
, '\0', nla_len(a
))) {
1306 OVS_NLERR(log
, "Invalid conntrack helper");
1310 #ifdef CONFIG_NF_NAT_NEEDED
1311 case OVS_CT_ATTR_NAT
: {
1312 int err
= parse_nat(a
, info
, log
);
1320 OVS_NLERR(log
, "Unknown conntrack attr (%d)",
1326 #ifdef CONFIG_NF_CONNTRACK_MARK
1327 if (!info
->commit
&& info
->mark
.mask
) {
1329 "Setting conntrack mark requires 'commit' flag.");
1333 #ifdef CONFIG_NF_CONNTRACK_LABELS
1334 if (!info
->commit
&& labels_nonzero(&info
->labels
.mask
)) {
1336 "Setting conntrack labels requires 'commit' flag.");
1341 OVS_NLERR(log
, "Conntrack attr has %d unknown bytes", rem
);
1348 bool ovs_ct_verify(struct net
*net
, enum ovs_key_attr attr
)
1350 if (attr
== OVS_KEY_ATTR_CT_STATE
)
1352 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1353 attr
== OVS_KEY_ATTR_CT_ZONE
)
1355 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) &&
1356 attr
== OVS_KEY_ATTR_CT_MARK
)
1358 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1359 attr
== OVS_KEY_ATTR_CT_LABELS
) {
1360 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1362 return ovs_net
->xt_label
;
1368 int ovs_ct_copy_action(struct net
*net
, const struct nlattr
*attr
,
1369 const struct sw_flow_key
*key
,
1370 struct sw_flow_actions
**sfa
, bool log
)
1372 struct ovs_conntrack_info ct_info
;
1373 const char *helper
= NULL
;
1377 family
= key_to_nfproto(key
);
1378 if (family
== NFPROTO_UNSPEC
) {
1379 OVS_NLERR(log
, "ct family unspecified");
1383 memset(&ct_info
, 0, sizeof(ct_info
));
1384 ct_info
.family
= family
;
1386 nf_ct_zone_init(&ct_info
.zone
, NF_CT_DEFAULT_ZONE_ID
,
1387 NF_CT_DEFAULT_ZONE_DIR
, 0);
1389 err
= parse_ct(attr
, &ct_info
, &helper
, log
);
1393 /* Set up template for tracking connections in specific zones. */
1394 ct_info
.ct
= nf_ct_tmpl_alloc(net
, &ct_info
.zone
, GFP_KERNEL
);
1396 OVS_NLERR(log
, "Failed to allocate conntrack template");
1400 __set_bit(IPS_CONFIRMED_BIT
, &ct_info
.ct
->status
);
1401 nf_conntrack_get(&ct_info
.ct
->ct_general
);
1404 err
= ovs_ct_add_helper(&ct_info
, helper
, key
, log
);
1409 err
= ovs_nla_add_action(sfa
, OVS_ACTION_ATTR_CT
, &ct_info
,
1410 sizeof(ct_info
), log
);
1416 __ovs_ct_free_action(&ct_info
);
1420 #ifdef CONFIG_NF_NAT_NEEDED
1421 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info
*info
,
1422 struct sk_buff
*skb
)
1424 struct nlattr
*start
;
1426 start
= nla_nest_start(skb
, OVS_CT_ATTR_NAT
);
1430 if (info
->nat
& OVS_CT_SRC_NAT
) {
1431 if (nla_put_flag(skb
, OVS_NAT_ATTR_SRC
))
1433 } else if (info
->nat
& OVS_CT_DST_NAT
) {
1434 if (nla_put_flag(skb
, OVS_NAT_ATTR_DST
))
1440 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
) {
1441 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
1442 info
->family
== NFPROTO_IPV4
) {
1443 if (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1444 info
->range
.min_addr
.ip
) ||
1445 (info
->range
.max_addr
.ip
1446 != info
->range
.min_addr
.ip
&&
1447 (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1448 info
->range
.max_addr
.ip
))))
1450 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
1451 info
->family
== NFPROTO_IPV6
) {
1452 if (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1453 &info
->range
.min_addr
.in6
) ||
1454 (memcmp(&info
->range
.max_addr
.in6
,
1455 &info
->range
.min_addr
.in6
,
1456 sizeof(info
->range
.max_addr
.in6
)) &&
1457 (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1458 &info
->range
.max_addr
.in6
))))
1464 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1465 (nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MIN
,
1466 ntohs(info
->range
.min_proto
.all
)) ||
1467 (info
->range
.max_proto
.all
!= info
->range
.min_proto
.all
&&
1468 nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MAX
,
1469 ntohs(info
->range
.max_proto
.all
)))))
1472 if (info
->range
.flags
& NF_NAT_RANGE_PERSISTENT
&&
1473 nla_put_flag(skb
, OVS_NAT_ATTR_PERSISTENT
))
1475 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM
&&
1476 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_HASH
))
1478 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM_FULLY
&&
1479 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_RANDOM
))
1482 nla_nest_end(skb
, start
);
1488 int ovs_ct_action_to_attr(const struct ovs_conntrack_info
*ct_info
,
1489 struct sk_buff
*skb
)
1491 struct nlattr
*start
;
1493 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_CT
);
1497 if (ct_info
->commit
&& nla_put_flag(skb
, ct_info
->force
1498 ? OVS_CT_ATTR_FORCE_COMMIT
1499 : OVS_CT_ATTR_COMMIT
))
1501 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1502 nla_put_u16(skb
, OVS_CT_ATTR_ZONE
, ct_info
->zone
.id
))
1504 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) && ct_info
->mark
.mask
&&
1505 nla_put(skb
, OVS_CT_ATTR_MARK
, sizeof(ct_info
->mark
),
1508 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1509 labels_nonzero(&ct_info
->labels
.mask
) &&
1510 nla_put(skb
, OVS_CT_ATTR_LABELS
, sizeof(ct_info
->labels
),
1513 if (ct_info
->helper
) {
1514 if (nla_put_string(skb
, OVS_CT_ATTR_HELPER
,
1515 ct_info
->helper
->name
))
1518 #ifdef CONFIG_NF_NAT_NEEDED
1519 if (ct_info
->nat
&& !ovs_ct_nat_to_attr(ct_info
, skb
))
1522 nla_nest_end(skb
, start
);
1527 void ovs_ct_free_action(const struct nlattr
*a
)
1529 struct ovs_conntrack_info
*ct_info
= nla_data(a
);
1531 __ovs_ct_free_action(ct_info
);
1534 static void __ovs_ct_free_action(struct ovs_conntrack_info
*ct_info
)
1536 if (ct_info
->helper
)
1537 module_put(ct_info
->helper
->me
);
1539 nf_ct_tmpl_free(ct_info
->ct
);
1542 void ovs_ct_init(struct net
*net
)
1544 unsigned int n_bits
= sizeof(struct ovs_key_ct_labels
) * BITS_PER_BYTE
;
1545 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1547 if (nf_connlabels_get(net
, n_bits
- 1)) {
1548 ovs_net
->xt_label
= false;
1549 OVS_NLERR(true, "Failed to set connlabel length");
1551 ovs_net
->xt_label
= true;
1555 void ovs_ct_exit(struct net
*net
)
1557 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1559 if (ovs_net
->xt_label
)
1560 nf_connlabels_put(net
);