2 * Copyright (c) 2015 Nicira, Inc.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
14 #include <linux/module.h>
15 #include <linux/openvswitch.h>
16 #include <linux/tcp.h>
17 #include <linux/udp.h>
18 #include <linux/sctp.h>
20 #include <net/netfilter/nf_conntrack_core.h>
21 #include <net/netfilter/nf_conntrack_helper.h>
22 #include <net/netfilter/nf_conntrack_labels.h>
23 #include <net/netfilter/nf_conntrack_seqadj.h>
24 #include <net/netfilter/nf_conntrack_zones.h>
25 #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
27 #ifdef CONFIG_NF_NAT_NEEDED
28 #include <linux/netfilter/nf_nat.h>
29 #include <net/netfilter/nf_nat_core.h>
30 #include <net/netfilter/nf_nat_l3proto.h>
34 #include "conntrack.h"
36 #include "flow_netlink.h"
38 struct ovs_ct_len_tbl
{
43 /* Metadata mark for masked write to conntrack mark */
49 /* Metadata label for masked write to conntrack label. */
51 struct ovs_key_ct_labels value
;
52 struct ovs_key_ct_labels mask
;
56 OVS_CT_NAT
= 1 << 0, /* NAT for committed connections only. */
57 OVS_CT_SRC_NAT
= 1 << 1, /* Source NAT for NEW connections. */
58 OVS_CT_DST_NAT
= 1 << 2, /* Destination NAT for NEW connections. */
61 /* Conntrack action context for execution. */
62 struct ovs_conntrack_info
{
63 struct nf_conntrack_helper
*helper
;
64 struct nf_conntrack_zone zone
;
67 u8 nat
: 3; /* enum ovs_ct_nat */
69 u8 have_eventmask
: 1;
71 u32 eventmask
; /* Mask of 1 << IPCT_*. */
73 struct md_labels labels
;
74 #ifdef CONFIG_NF_NAT_NEEDED
75 struct nf_nat_range range
; /* Only present for SRC NAT and DST NAT. */
79 static bool labels_nonzero(const struct ovs_key_ct_labels
*labels
);
81 static void __ovs_ct_free_action(struct ovs_conntrack_info
*ct_info
);
83 static u16
key_to_nfproto(const struct sw_flow_key
*key
)
85 switch (ntohs(key
->eth
.type
)) {
91 return NFPROTO_UNSPEC
;
95 /* Map SKB connection state into the values used by flow definition. */
96 static u8
ovs_ct_get_state(enum ip_conntrack_info ctinfo
)
98 u8 ct_state
= OVS_CS_F_TRACKED
;
101 case IP_CT_ESTABLISHED_REPLY
:
102 case IP_CT_RELATED_REPLY
:
103 ct_state
|= OVS_CS_F_REPLY_DIR
;
110 case IP_CT_ESTABLISHED
:
111 case IP_CT_ESTABLISHED_REPLY
:
112 ct_state
|= OVS_CS_F_ESTABLISHED
;
115 case IP_CT_RELATED_REPLY
:
116 ct_state
|= OVS_CS_F_RELATED
;
119 ct_state
|= OVS_CS_F_NEW
;
128 static u32
ovs_ct_get_mark(const struct nf_conn
*ct
)
130 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
131 return ct
? ct
->mark
: 0;
137 /* Guard against conntrack labels max size shrinking below 128 bits. */
138 #if NF_CT_LABELS_MAX_SIZE < 16
139 #error NF_CT_LABELS_MAX_SIZE must be at least 16 bytes
142 static void ovs_ct_get_labels(const struct nf_conn
*ct
,
143 struct ovs_key_ct_labels
*labels
)
145 struct nf_conn_labels
*cl
= ct
? nf_ct_labels_find(ct
) : NULL
;
148 memcpy(labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
150 memset(labels
, 0, OVS_CT_LABELS_LEN
);
153 static void __ovs_ct_update_key_orig_tp(struct sw_flow_key
*key
,
154 const struct nf_conntrack_tuple
*orig
,
157 key
->ct_orig_proto
= orig
->dst
.protonum
;
158 if (orig
->dst
.protonum
== icmp_proto
) {
159 key
->ct
.orig_tp
.src
= htons(orig
->dst
.u
.icmp
.type
);
160 key
->ct
.orig_tp
.dst
= htons(orig
->dst
.u
.icmp
.code
);
162 key
->ct
.orig_tp
.src
= orig
->src
.u
.all
;
163 key
->ct
.orig_tp
.dst
= orig
->dst
.u
.all
;
167 static void __ovs_ct_update_key(struct sw_flow_key
*key
, u8 state
,
168 const struct nf_conntrack_zone
*zone
,
169 const struct nf_conn
*ct
)
171 key
->ct_state
= state
;
172 key
->ct_zone
= zone
->id
;
173 key
->ct
.mark
= ovs_ct_get_mark(ct
);
174 ovs_ct_get_labels(ct
, &key
->ct
.labels
);
177 const struct nf_conntrack_tuple
*orig
;
179 /* Use the master if we have one. */
182 orig
= &ct
->tuplehash
[IP_CT_DIR_ORIGINAL
].tuple
;
184 /* IP version must match with the master connection. */
185 if (key
->eth
.type
== htons(ETH_P_IP
) &&
186 nf_ct_l3num(ct
) == NFPROTO_IPV4
) {
187 key
->ipv4
.ct_orig
.src
= orig
->src
.u3
.ip
;
188 key
->ipv4
.ct_orig
.dst
= orig
->dst
.u3
.ip
;
189 __ovs_ct_update_key_orig_tp(key
, orig
, IPPROTO_ICMP
);
191 } else if (key
->eth
.type
== htons(ETH_P_IPV6
) &&
192 !sw_flow_key_is_nd(key
) &&
193 nf_ct_l3num(ct
) == NFPROTO_IPV6
) {
194 key
->ipv6
.ct_orig
.src
= orig
->src
.u3
.in6
;
195 key
->ipv6
.ct_orig
.dst
= orig
->dst
.u3
.in6
;
196 __ovs_ct_update_key_orig_tp(key
, orig
, NEXTHDR_ICMP
);
200 /* Clear 'ct_orig_proto' to mark the non-existence of conntrack
201 * original direction key fields.
203 key
->ct_orig_proto
= 0;
206 /* Update 'key' based on skb->_nfct. If 'post_ct' is true, then OVS has
207 * previously sent the packet to conntrack via the ct action. If
208 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
209 * initialized from the connection status.
211 static void ovs_ct_update_key(const struct sk_buff
*skb
,
212 const struct ovs_conntrack_info
*info
,
213 struct sw_flow_key
*key
, bool post_ct
,
216 const struct nf_conntrack_zone
*zone
= &nf_ct_zone_dflt
;
217 enum ip_conntrack_info ctinfo
;
221 ct
= nf_ct_get(skb
, &ctinfo
);
223 state
= ovs_ct_get_state(ctinfo
);
224 /* All unconfirmed entries are NEW connections. */
225 if (!nf_ct_is_confirmed(ct
))
226 state
|= OVS_CS_F_NEW
;
227 /* OVS persists the related flag for the duration of the
231 state
|= OVS_CS_F_RELATED
;
232 if (keep_nat_flags
) {
233 state
|= key
->ct_state
& OVS_CS_F_NAT_MASK
;
235 if (ct
->status
& IPS_SRC_NAT
)
236 state
|= OVS_CS_F_SRC_NAT
;
237 if (ct
->status
& IPS_DST_NAT
)
238 state
|= OVS_CS_F_DST_NAT
;
240 zone
= nf_ct_zone(ct
);
241 } else if (post_ct
) {
242 state
= OVS_CS_F_TRACKED
| OVS_CS_F_INVALID
;
246 __ovs_ct_update_key(key
, state
, zone
, ct
);
249 /* This is called to initialize CT key fields possibly coming in from the local
252 void ovs_ct_fill_key(const struct sk_buff
*skb
, struct sw_flow_key
*key
)
254 ovs_ct_update_key(skb
, NULL
, key
, false, false);
257 #define IN6_ADDR_INITIALIZER(ADDR) \
258 { (ADDR).s6_addr32[0], (ADDR).s6_addr32[1], \
259 (ADDR).s6_addr32[2], (ADDR).s6_addr32[3] }
261 int ovs_ct_put_key(const struct sw_flow_key
*swkey
,
262 const struct sw_flow_key
*output
, struct sk_buff
*skb
)
264 if (nla_put_u32(skb
, OVS_KEY_ATTR_CT_STATE
, output
->ct_state
))
267 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
268 nla_put_u16(skb
, OVS_KEY_ATTR_CT_ZONE
, output
->ct_zone
))
271 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) &&
272 nla_put_u32(skb
, OVS_KEY_ATTR_CT_MARK
, output
->ct
.mark
))
275 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
276 nla_put(skb
, OVS_KEY_ATTR_CT_LABELS
, sizeof(output
->ct
.labels
),
280 if (swkey
->ct_orig_proto
) {
281 if (swkey
->eth
.type
== htons(ETH_P_IP
)) {
282 struct ovs_key_ct_tuple_ipv4 orig
= {
283 output
->ipv4
.ct_orig
.src
,
284 output
->ipv4
.ct_orig
.dst
,
285 output
->ct
.orig_tp
.src
,
286 output
->ct
.orig_tp
.dst
,
287 output
->ct_orig_proto
,
289 if (nla_put(skb
, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV4
,
290 sizeof(orig
), &orig
))
292 } else if (swkey
->eth
.type
== htons(ETH_P_IPV6
)) {
293 struct ovs_key_ct_tuple_ipv6 orig
= {
294 IN6_ADDR_INITIALIZER(output
->ipv6
.ct_orig
.src
),
295 IN6_ADDR_INITIALIZER(output
->ipv6
.ct_orig
.dst
),
296 output
->ct
.orig_tp
.src
,
297 output
->ct
.orig_tp
.dst
,
298 output
->ct_orig_proto
,
300 if (nla_put(skb
, OVS_KEY_ATTR_CT_ORIG_TUPLE_IPV6
,
301 sizeof(orig
), &orig
))
309 static int ovs_ct_set_mark(struct nf_conn
*ct
, struct sw_flow_key
*key
,
310 u32 ct_mark
, u32 mask
)
312 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
315 new_mark
= ct_mark
| (ct
->mark
& ~(mask
));
316 if (ct
->mark
!= new_mark
) {
318 if (nf_ct_is_confirmed(ct
))
319 nf_conntrack_event_cache(IPCT_MARK
, ct
);
320 key
->ct
.mark
= new_mark
;
329 static struct nf_conn_labels
*ovs_ct_get_conn_labels(struct nf_conn
*ct
)
331 struct nf_conn_labels
*cl
;
333 cl
= nf_ct_labels_find(ct
);
335 nf_ct_labels_ext_add(ct
);
336 cl
= nf_ct_labels_find(ct
);
342 /* Initialize labels for a new, yet to be committed conntrack entry. Note that
343 * since the new connection is not yet confirmed, and thus no-one else has
344 * access to it's labels, we simply write them over.
346 static int ovs_ct_init_labels(struct nf_conn
*ct
, struct sw_flow_key
*key
,
347 const struct ovs_key_ct_labels
*labels
,
348 const struct ovs_key_ct_labels
*mask
)
350 struct nf_conn_labels
*cl
, *master_cl
;
351 bool have_mask
= labels_nonzero(mask
);
353 /* Inherit master's labels to the related connection? */
354 master_cl
= ct
->master
? nf_ct_labels_find(ct
->master
) : NULL
;
356 if (!master_cl
&& !have_mask
)
357 return 0; /* Nothing to do. */
359 cl
= ovs_ct_get_conn_labels(ct
);
363 /* Inherit the master's labels, if any. */
368 u32
*dst
= (u32
*)cl
->bits
;
371 for (i
= 0; i
< OVS_CT_LABELS_LEN_32
; i
++)
372 dst
[i
] = (dst
[i
] & ~mask
->ct_labels_32
[i
]) |
373 (labels
->ct_labels_32
[i
]
374 & mask
->ct_labels_32
[i
]);
377 /* Labels are included in the IPCTNL_MSG_CT_NEW event only if the
378 * IPCT_LABEL bit is set in the event cache.
380 nf_conntrack_event_cache(IPCT_LABEL
, ct
);
382 memcpy(&key
->ct
.labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
387 static int ovs_ct_set_labels(struct nf_conn
*ct
, struct sw_flow_key
*key
,
388 const struct ovs_key_ct_labels
*labels
,
389 const struct ovs_key_ct_labels
*mask
)
391 struct nf_conn_labels
*cl
;
394 cl
= ovs_ct_get_conn_labels(ct
);
398 err
= nf_connlabels_replace(ct
, labels
->ct_labels_32
,
400 OVS_CT_LABELS_LEN_32
);
404 memcpy(&key
->ct
.labels
, cl
->bits
, OVS_CT_LABELS_LEN
);
409 /* 'skb' should already be pulled to nh_ofs. */
410 static int ovs_ct_helper(struct sk_buff
*skb
, u16 proto
)
412 const struct nf_conntrack_helper
*helper
;
413 const struct nf_conn_help
*help
;
414 enum ip_conntrack_info ctinfo
;
415 unsigned int protoff
;
419 ct
= nf_ct_get(skb
, &ctinfo
);
420 if (!ct
|| ctinfo
== IP_CT_RELATED_REPLY
)
423 help
= nfct_help(ct
);
427 helper
= rcu_dereference(help
->helper
);
433 protoff
= ip_hdrlen(skb
);
436 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
440 ofs
= ipv6_skip_exthdr(skb
, sizeof(struct ipv6hdr
), &nexthdr
,
442 if (ofs
< 0 || (frag_off
& htons(~0x7)) != 0) {
443 pr_debug("proto header not found\n");
450 WARN_ONCE(1, "helper invoked on non-IP family!");
454 err
= helper
->help(skb
, protoff
, ct
, ctinfo
);
455 if (err
!= NF_ACCEPT
)
458 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
459 * FTP with NAT) adusting the TCP payload size when mangling IP
460 * addresses and/or port numbers in the text-based control connection.
462 if (test_bit(IPS_SEQ_ADJUST_BIT
, &ct
->status
) &&
463 !nf_ct_seq_adjust(skb
, ct
, ctinfo
, protoff
))
468 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
469 * value if 'skb' is freed.
471 static int handle_fragments(struct net
*net
, struct sw_flow_key
*key
,
472 u16 zone
, struct sk_buff
*skb
)
474 struct ovs_skb_cb ovs_cb
= *OVS_CB(skb
);
477 if (key
->eth
.type
== htons(ETH_P_IP
)) {
478 enum ip_defrag_users user
= IP_DEFRAG_CONNTRACK_IN
+ zone
;
480 memset(IPCB(skb
), 0, sizeof(struct inet_skb_parm
));
481 err
= ip_defrag(net
, skb
, user
);
485 ovs_cb
.mru
= IPCB(skb
)->frag_max_size
;
486 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
487 } else if (key
->eth
.type
== htons(ETH_P_IPV6
)) {
488 enum ip6_defrag_users user
= IP6_DEFRAG_CONNTRACK_IN
+ zone
;
490 memset(IP6CB(skb
), 0, sizeof(struct inet6_skb_parm
));
491 err
= nf_ct_frag6_gather(net
, skb
, user
);
493 if (err
!= -EINPROGRESS
)
498 key
->ip
.proto
= ipv6_hdr(skb
)->nexthdr
;
499 ovs_cb
.mru
= IP6CB(skb
)->frag_max_size
;
503 return -EPFNOSUPPORT
;
506 key
->ip
.frag
= OVS_FRAG_TYPE_NONE
;
509 *OVS_CB(skb
) = ovs_cb
;
514 static struct nf_conntrack_expect
*
515 ovs_ct_expect_find(struct net
*net
, const struct nf_conntrack_zone
*zone
,
516 u16 proto
, const struct sk_buff
*skb
)
518 struct nf_conntrack_tuple tuple
;
519 struct nf_conntrack_expect
*exp
;
521 if (!nf_ct_get_tuplepr(skb
, skb_network_offset(skb
), proto
, net
, &tuple
))
524 exp
= __nf_ct_expect_find(net
, zone
, &tuple
);
526 struct nf_conntrack_tuple_hash
*h
;
528 /* Delete existing conntrack entry, if it clashes with the
529 * expectation. This can happen since conntrack ALGs do not
530 * check for clashes between (new) expectations and existing
531 * conntrack entries. nf_conntrack_in() will check the
532 * expectations only if a conntrack entry can not be found,
533 * which can lead to OVS finding the expectation (here) in the
534 * init direction, but which will not be removed by the
535 * nf_conntrack_in() call, if a matching conntrack entry is
536 * found instead. In this case all init direction packets
537 * would be reported as new related packets, while reply
538 * direction packets would be reported as un-related
539 * established packets.
541 h
= nf_conntrack_find_get(net
, zone
, &tuple
);
543 struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(h
);
545 nf_ct_delete(ct
, 0, 0);
546 nf_conntrack_put(&ct
->ct_general
);
553 /* This replicates logic from nf_conntrack_core.c that is not exported. */
554 static enum ip_conntrack_info
555 ovs_ct_get_info(const struct nf_conntrack_tuple_hash
*h
)
557 const struct nf_conn
*ct
= nf_ct_tuplehash_to_ctrack(h
);
559 if (NF_CT_DIRECTION(h
) == IP_CT_DIR_REPLY
)
560 return IP_CT_ESTABLISHED_REPLY
;
561 /* Once we've had two way comms, always ESTABLISHED. */
562 if (test_bit(IPS_SEEN_REPLY_BIT
, &ct
->status
))
563 return IP_CT_ESTABLISHED
;
564 if (test_bit(IPS_EXPECTED_BIT
, &ct
->status
))
565 return IP_CT_RELATED
;
569 /* Find an existing connection which this packet belongs to without
570 * re-attributing statistics or modifying the connection state. This allows an
571 * skb->_nfct lost due to an upcall to be recovered during actions execution.
573 * Must be called with rcu_read_lock.
575 * On success, populates skb->_nfct and returns the connection. Returns NULL
576 * if there is no existing entry.
578 static struct nf_conn
*
579 ovs_ct_find_existing(struct net
*net
, const struct nf_conntrack_zone
*zone
,
580 u8 l3num
, struct sk_buff
*skb
, bool natted
)
582 const struct nf_conntrack_l3proto
*l3proto
;
583 const struct nf_conntrack_l4proto
*l4proto
;
584 struct nf_conntrack_tuple tuple
;
585 struct nf_conntrack_tuple_hash
*h
;
587 unsigned int dataoff
;
590 l3proto
= __nf_ct_l3proto_find(l3num
);
591 if (l3proto
->get_l4proto(skb
, skb_network_offset(skb
), &dataoff
,
593 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
596 l4proto
= __nf_ct_l4proto_find(l3num
, protonum
);
597 if (!nf_ct_get_tuple(skb
, skb_network_offset(skb
), dataoff
, l3num
,
598 protonum
, net
, &tuple
, l3proto
, l4proto
)) {
599 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
603 /* Must invert the tuple if skb has been transformed by NAT. */
605 struct nf_conntrack_tuple inverse
;
607 if (!nf_ct_invert_tuple(&inverse
, &tuple
, l3proto
, l4proto
)) {
608 pr_debug("ovs_ct_find_existing: Inversion failed!\n");
614 /* look for tuple match */
615 h
= nf_conntrack_find_get(net
, zone
, &tuple
);
617 return NULL
; /* Not found. */
619 ct
= nf_ct_tuplehash_to_ctrack(h
);
621 /* Inverted packet tuple matches the reverse direction conntrack tuple,
622 * select the other tuplehash to get the right 'ctinfo' bits for this
626 h
= &ct
->tuplehash
[!h
->tuple
.dst
.dir
];
628 nf_ct_set(skb
, ct
, ovs_ct_get_info(h
));
633 struct nf_conn
*ovs_ct_executed(struct net
*net
,
634 const struct sw_flow_key
*key
,
635 const struct ovs_conntrack_info
*info
,
639 struct nf_conn
*ct
= NULL
;
641 /* If no ct, check if we have evidence that an existing conntrack entry
642 * might be found for this skb. This happens when we lose a skb->_nfct
643 * due to an upcall, or if the direction is being forced. If the
644 * connection was not confirmed, it is not cached and needs to be run
645 * through conntrack again.
647 *ct_executed
= (key
->ct_state
& OVS_CS_F_TRACKED
) &&
648 !(key
->ct_state
& OVS_CS_F_INVALID
) &&
649 (key
->ct_zone
== info
->zone
.id
);
651 if (*ct_executed
|| (!key
->ct_state
&& info
->force
)) {
652 ct
= ovs_ct_find_existing(net
, &info
->zone
, info
->family
, skb
,
660 /* Determine whether skb->_nfct is equal to the result of conntrack lookup. */
661 static bool skb_nfct_cached(struct net
*net
,
662 const struct sw_flow_key
*key
,
663 const struct ovs_conntrack_info
*info
,
666 enum ip_conntrack_info ctinfo
;
668 bool ct_executed
= true;
670 ct
= nf_ct_get(skb
, &ctinfo
);
672 ct
= ovs_ct_executed(net
, key
, info
, skb
, &ct_executed
);
675 nf_ct_get(skb
, &ctinfo
);
679 if (!net_eq(net
, read_pnet(&ct
->ct_net
)))
681 if (!nf_ct_zone_equal_any(info
->ct
, nf_ct_zone(ct
)))
684 struct nf_conn_help
*help
;
686 help
= nf_ct_ext_find(ct
, NF_CT_EXT_HELPER
);
687 if (help
&& rcu_access_pointer(help
->helper
) != info
->helper
)
690 /* Force conntrack entry direction to the current packet? */
691 if (info
->force
&& CTINFO2DIR(ctinfo
) != IP_CT_DIR_ORIGINAL
) {
692 /* Delete the conntrack entry if confirmed, else just release
695 if (nf_ct_is_confirmed(ct
))
696 nf_ct_delete(ct
, 0, 0);
698 nf_conntrack_put(&ct
->ct_general
);
699 nf_ct_set(skb
, NULL
, 0);
706 #ifdef CONFIG_NF_NAT_NEEDED
707 /* Modelled after nf_nat_ipv[46]_fn().
708 * range is only used for new, uninitialized NAT state.
709 * Returns either NF_ACCEPT or NF_DROP.
711 static int ovs_ct_nat_execute(struct sk_buff
*skb
, struct nf_conn
*ct
,
712 enum ip_conntrack_info ctinfo
,
713 const struct nf_nat_range
*range
,
714 enum nf_nat_manip_type maniptype
)
716 int hooknum
, nh_off
, err
= NF_ACCEPT
;
718 nh_off
= skb_network_offset(skb
);
719 skb_pull_rcsum(skb
, nh_off
);
721 /* See HOOK2MANIP(). */
722 if (maniptype
== NF_NAT_MANIP_SRC
)
723 hooknum
= NF_INET_LOCAL_IN
; /* Source NAT */
725 hooknum
= NF_INET_LOCAL_OUT
; /* Destination NAT */
729 case IP_CT_RELATED_REPLY
:
730 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
731 skb
->protocol
== htons(ETH_P_IP
) &&
732 ip_hdr(skb
)->protocol
== IPPROTO_ICMP
) {
733 if (!nf_nat_icmp_reply_translation(skb
, ct
, ctinfo
,
737 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
738 skb
->protocol
== htons(ETH_P_IPV6
)) {
740 u8 nexthdr
= ipv6_hdr(skb
)->nexthdr
;
741 int hdrlen
= ipv6_skip_exthdr(skb
,
742 sizeof(struct ipv6hdr
),
743 &nexthdr
, &frag_off
);
745 if (hdrlen
>= 0 && nexthdr
== IPPROTO_ICMPV6
) {
746 if (!nf_nat_icmpv6_reply_translation(skb
, ct
,
754 /* Non-ICMP, fall thru to initialize if needed. */
756 /* Seen it before? This can happen for loopback, retrans,
759 if (!nf_nat_initialized(ct
, maniptype
)) {
760 /* Initialize according to the NAT action. */
761 err
= (range
&& range
->flags
& NF_NAT_RANGE_MAP_IPS
)
762 /* Action is set up to establish a new
765 ? nf_nat_setup_info(ct
, range
, maniptype
)
766 : nf_nat_alloc_null_binding(ct
, hooknum
);
767 if (err
!= NF_ACCEPT
)
772 case IP_CT_ESTABLISHED
:
773 case IP_CT_ESTABLISHED_REPLY
:
781 err
= nf_nat_packet(ct
, ctinfo
, hooknum
, skb
);
783 skb_push(skb
, nh_off
);
784 skb_postpush_rcsum(skb
, skb
->data
, nh_off
);
789 static void ovs_nat_update_key(struct sw_flow_key
*key
,
790 const struct sk_buff
*skb
,
791 enum nf_nat_manip_type maniptype
)
793 if (maniptype
== NF_NAT_MANIP_SRC
) {
796 key
->ct_state
|= OVS_CS_F_SRC_NAT
;
797 if (key
->eth
.type
== htons(ETH_P_IP
))
798 key
->ipv4
.addr
.src
= ip_hdr(skb
)->saddr
;
799 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
800 memcpy(&key
->ipv6
.addr
.src
, &ipv6_hdr(skb
)->saddr
,
801 sizeof(key
->ipv6
.addr
.src
));
805 if (key
->ip
.proto
== IPPROTO_UDP
)
806 src
= udp_hdr(skb
)->source
;
807 else if (key
->ip
.proto
== IPPROTO_TCP
)
808 src
= tcp_hdr(skb
)->source
;
809 else if (key
->ip
.proto
== IPPROTO_SCTP
)
810 src
= sctp_hdr(skb
)->source
;
818 key
->ct_state
|= OVS_CS_F_DST_NAT
;
819 if (key
->eth
.type
== htons(ETH_P_IP
))
820 key
->ipv4
.addr
.dst
= ip_hdr(skb
)->daddr
;
821 else if (key
->eth
.type
== htons(ETH_P_IPV6
))
822 memcpy(&key
->ipv6
.addr
.dst
, &ipv6_hdr(skb
)->daddr
,
823 sizeof(key
->ipv6
.addr
.dst
));
827 if (key
->ip
.proto
== IPPROTO_UDP
)
828 dst
= udp_hdr(skb
)->dest
;
829 else if (key
->ip
.proto
== IPPROTO_TCP
)
830 dst
= tcp_hdr(skb
)->dest
;
831 else if (key
->ip
.proto
== IPPROTO_SCTP
)
832 dst
= sctp_hdr(skb
)->dest
;
840 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
841 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
842 const struct ovs_conntrack_info
*info
,
843 struct sk_buff
*skb
, struct nf_conn
*ct
,
844 enum ip_conntrack_info ctinfo
)
846 enum nf_nat_manip_type maniptype
;
849 /* Add NAT extension if not confirmed yet. */
850 if (!nf_ct_is_confirmed(ct
) && !nf_ct_nat_ext_add(ct
))
851 return NF_ACCEPT
; /* Can't NAT. */
853 /* Determine NAT type.
854 * Check if the NAT type can be deduced from the tracked connection.
855 * Make sure new expected connections (IP_CT_RELATED) are NATted only
858 if (info
->nat
& OVS_CT_NAT
&& ctinfo
!= IP_CT_NEW
&&
859 ct
->status
& IPS_NAT_MASK
&&
860 (ctinfo
!= IP_CT_RELATED
|| info
->commit
)) {
861 /* NAT an established or related connection like before. */
862 if (CTINFO2DIR(ctinfo
) == IP_CT_DIR_REPLY
)
863 /* This is the REPLY direction for a connection
864 * for which NAT was applied in the forward
865 * direction. Do the reverse NAT.
867 maniptype
= ct
->status
& IPS_SRC_NAT
868 ? NF_NAT_MANIP_DST
: NF_NAT_MANIP_SRC
;
870 maniptype
= ct
->status
& IPS_SRC_NAT
871 ? NF_NAT_MANIP_SRC
: NF_NAT_MANIP_DST
;
872 } else if (info
->nat
& OVS_CT_SRC_NAT
) {
873 maniptype
= NF_NAT_MANIP_SRC
;
874 } else if (info
->nat
& OVS_CT_DST_NAT
) {
875 maniptype
= NF_NAT_MANIP_DST
;
877 return NF_ACCEPT
; /* Connection is not NATed. */
879 err
= ovs_ct_nat_execute(skb
, ct
, ctinfo
, &info
->range
, maniptype
);
881 /* Mark NAT done if successful and update the flow key. */
882 if (err
== NF_ACCEPT
)
883 ovs_nat_update_key(key
, skb
, maniptype
);
887 #else /* !CONFIG_NF_NAT_NEEDED */
888 static int ovs_ct_nat(struct net
*net
, struct sw_flow_key
*key
,
889 const struct ovs_conntrack_info
*info
,
890 struct sk_buff
*skb
, struct nf_conn
*ct
,
891 enum ip_conntrack_info ctinfo
)
897 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
898 * not done already. Update key with new CT state after passing the packet
900 * Note that if the packet is deemed invalid by conntrack, skb->_nfct will be
901 * set to NULL and 0 will be returned.
903 static int __ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
904 const struct ovs_conntrack_info
*info
,
907 /* If we are recirculating packets to match on conntrack fields and
908 * committing with a separate conntrack action, then we don't need to
909 * actually run the packet through conntrack twice unless it's for a
912 bool cached
= skb_nfct_cached(net
, key
, info
, skb
);
913 enum ip_conntrack_info ctinfo
;
917 struct nf_conn
*tmpl
= info
->ct
;
920 /* Associate skb with specified zone. */
923 nf_conntrack_put(skb_nfct(skb
));
924 nf_conntrack_get(&tmpl
->ct_general
);
925 nf_ct_set(skb
, tmpl
, IP_CT_NEW
);
928 err
= nf_conntrack_in(net
, info
->family
,
929 NF_INET_PRE_ROUTING
, skb
);
930 if (err
!= NF_ACCEPT
)
933 /* Clear CT state NAT flags to mark that we have not yet done
934 * NAT after the nf_conntrack_in() call. We can actually clear
935 * the whole state, as it will be re-initialized below.
939 /* Update the key, but keep the NAT flags. */
940 ovs_ct_update_key(skb
, info
, key
, true, true);
943 ct
= nf_ct_get(skb
, &ctinfo
);
945 /* Packets starting a new connection must be NATted before the
946 * helper, so that the helper knows about the NAT. We enforce
947 * this by delaying both NAT and helper calls for unconfirmed
948 * connections until the committing CT action. For later
949 * packets NAT and Helper may be called in either order.
951 * NAT will be done only if the CT action has NAT, and only
952 * once per packet (per zone), as guarded by the NAT bits in
955 if (info
->nat
&& !(key
->ct_state
& OVS_CS_F_NAT_MASK
) &&
956 (nf_ct_is_confirmed(ct
) || info
->commit
) &&
957 ovs_ct_nat(net
, key
, info
, skb
, ct
, ctinfo
) != NF_ACCEPT
) {
961 /* Userspace may decide to perform a ct lookup without a helper
962 * specified followed by a (recirculate and) commit with one.
963 * Therefore, for unconfirmed connections which we will commit,
964 * we need to attach the helper here.
966 if (!nf_ct_is_confirmed(ct
) && info
->commit
&&
967 info
->helper
&& !nfct_help(ct
)) {
968 int err
= __nf_ct_try_assign_helper(ct
, info
->ct
,
974 /* Call the helper only if:
975 * - nf_conntrack_in() was executed above ("!cached") for a
976 * confirmed connection, or
977 * - When committing an unconfirmed connection.
979 if ((nf_ct_is_confirmed(ct
) ? !cached
: info
->commit
) &&
980 ovs_ct_helper(skb
, info
->family
) != NF_ACCEPT
) {
988 /* Lookup connection and read fields into key. */
989 static int ovs_ct_lookup(struct net
*net
, struct sw_flow_key
*key
,
990 const struct ovs_conntrack_info
*info
,
993 struct nf_conntrack_expect
*exp
;
995 /* If we pass an expected packet through nf_conntrack_in() the
996 * expectation is typically removed, but the packet could still be
997 * lost in upcall processing. To prevent this from happening we
998 * perform an explicit expectation lookup. Expected connections are
999 * always new, and will be passed through conntrack only when they are
1000 * committed, as it is OK to remove the expectation at that time.
1002 exp
= ovs_ct_expect_find(net
, &info
->zone
, info
->family
, skb
);
1006 /* NOTE: New connections are NATted and Helped only when
1007 * committed, so we are not calling into NAT here.
1009 state
= OVS_CS_F_TRACKED
| OVS_CS_F_NEW
| OVS_CS_F_RELATED
;
1010 __ovs_ct_update_key(key
, state
, &info
->zone
, exp
->master
);
1015 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
1019 ct
= (struct nf_conn
*)skb_nfct(skb
);
1021 nf_ct_deliver_cached_events(ct
);
1027 static bool labels_nonzero(const struct ovs_key_ct_labels
*labels
)
1031 for (i
= 0; i
< OVS_CT_LABELS_LEN_32
; i
++)
1032 if (labels
->ct_labels_32
[i
])
1038 /* Lookup connection and confirm if unconfirmed. */
1039 static int ovs_ct_commit(struct net
*net
, struct sw_flow_key
*key
,
1040 const struct ovs_conntrack_info
*info
,
1041 struct sk_buff
*skb
)
1043 enum ip_conntrack_info ctinfo
;
1047 err
= __ovs_ct_lookup(net
, key
, info
, skb
);
1051 /* The connection could be invalid, in which case this is a no-op.*/
1052 ct
= nf_ct_get(skb
, &ctinfo
);
1056 /* Set the conntrack event mask if given. NEW and DELETE events have
1057 * their own groups, but the NFNLGRP_CONNTRACK_UPDATE group listener
1058 * typically would receive many kinds of updates. Setting the event
1059 * mask allows those events to be filtered. The set event mask will
1060 * remain in effect for the lifetime of the connection unless changed
1061 * by a further CT action with both the commit flag and the eventmask
1063 if (info
->have_eventmask
) {
1064 struct nf_conntrack_ecache
*cache
= nf_ct_ecache_find(ct
);
1067 cache
->ctmask
= info
->eventmask
;
1070 /* Apply changes before confirming the connection so that the initial
1071 * conntrack NEW netlink event carries the values given in the CT
1074 if (info
->mark
.mask
) {
1075 err
= ovs_ct_set_mark(ct
, key
, info
->mark
.value
,
1080 if (!nf_ct_is_confirmed(ct
)) {
1081 err
= ovs_ct_init_labels(ct
, key
, &info
->labels
.value
,
1082 &info
->labels
.mask
);
1085 } else if (labels_nonzero(&info
->labels
.mask
)) {
1086 err
= ovs_ct_set_labels(ct
, key
, &info
->labels
.value
,
1087 &info
->labels
.mask
);
1091 /* This will take care of sending queued events even if the connection
1092 * is already confirmed.
1094 if (nf_conntrack_confirm(skb
) != NF_ACCEPT
)
1100 /* Trim the skb to the length specified by the IP/IPv6 header,
1101 * removing any trailing lower-layer padding. This prepares the skb
1102 * for higher-layer processing that assumes skb->len excludes padding
1103 * (such as nf_ip_checksum). The caller needs to pull the skb to the
1104 * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
1106 static int ovs_skb_network_trim(struct sk_buff
*skb
)
1111 switch (skb
->protocol
) {
1112 case htons(ETH_P_IP
):
1113 len
= ntohs(ip_hdr(skb
)->tot_len
);
1115 case htons(ETH_P_IPV6
):
1116 len
= sizeof(struct ipv6hdr
)
1117 + ntohs(ipv6_hdr(skb
)->payload_len
);
1123 err
= pskb_trim_rcsum(skb
, len
);
1130 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
1131 * value if 'skb' is freed.
1133 int ovs_ct_execute(struct net
*net
, struct sk_buff
*skb
,
1134 struct sw_flow_key
*key
,
1135 const struct ovs_conntrack_info
*info
)
1140 /* The conntrack module expects to be working at L3. */
1141 nh_ofs
= skb_network_offset(skb
);
1142 skb_pull_rcsum(skb
, nh_ofs
);
1144 err
= ovs_skb_network_trim(skb
);
1148 if (key
->ip
.frag
!= OVS_FRAG_TYPE_NONE
) {
1149 err
= handle_fragments(net
, key
, info
->zone
.id
, skb
);
1155 err
= ovs_ct_commit(net
, key
, info
, skb
);
1157 err
= ovs_ct_lookup(net
, key
, info
, skb
);
1159 skb_push(skb
, nh_ofs
);
1160 skb_postpush_rcsum(skb
, skb
->data
, nh_ofs
);
1166 static int ovs_ct_add_helper(struct ovs_conntrack_info
*info
, const char *name
,
1167 const struct sw_flow_key
*key
, bool log
)
1169 struct nf_conntrack_helper
*helper
;
1170 struct nf_conn_help
*help
;
1172 helper
= nf_conntrack_helper_try_module_get(name
, info
->family
,
1175 OVS_NLERR(log
, "Unknown helper \"%s\"", name
);
1179 help
= nf_ct_helper_ext_add(info
->ct
, helper
, GFP_KERNEL
);
1181 nf_conntrack_helper_put(helper
);
1185 rcu_assign_pointer(help
->helper
, helper
);
1186 info
->helper
= helper
;
1190 #ifdef CONFIG_NF_NAT_NEEDED
1191 static int parse_nat(const struct nlattr
*attr
,
1192 struct ovs_conntrack_info
*info
, bool log
)
1196 bool have_ip_max
= false;
1197 bool have_proto_max
= false;
1198 bool ip_vers
= (info
->family
== NFPROTO_IPV6
);
1200 nla_for_each_nested(a
, attr
, rem
) {
1201 static const int ovs_nat_attr_lens
[OVS_NAT_ATTR_MAX
+ 1][2] = {
1202 [OVS_NAT_ATTR_SRC
] = {0, 0},
1203 [OVS_NAT_ATTR_DST
] = {0, 0},
1204 [OVS_NAT_ATTR_IP_MIN
] = {sizeof(struct in_addr
),
1205 sizeof(struct in6_addr
)},
1206 [OVS_NAT_ATTR_IP_MAX
] = {sizeof(struct in_addr
),
1207 sizeof(struct in6_addr
)},
1208 [OVS_NAT_ATTR_PROTO_MIN
] = {sizeof(u16
), sizeof(u16
)},
1209 [OVS_NAT_ATTR_PROTO_MAX
] = {sizeof(u16
), sizeof(u16
)},
1210 [OVS_NAT_ATTR_PERSISTENT
] = {0, 0},
1211 [OVS_NAT_ATTR_PROTO_HASH
] = {0, 0},
1212 [OVS_NAT_ATTR_PROTO_RANDOM
] = {0, 0},
1214 int type
= nla_type(a
);
1216 if (type
> OVS_NAT_ATTR_MAX
) {
1217 OVS_NLERR(log
, "Unknown NAT attribute (type=%d, max=%d)",
1218 type
, OVS_NAT_ATTR_MAX
);
1222 if (nla_len(a
) != ovs_nat_attr_lens
[type
][ip_vers
]) {
1223 OVS_NLERR(log
, "NAT attribute type %d has unexpected length (%d != %d)",
1225 ovs_nat_attr_lens
[type
][ip_vers
]);
1230 case OVS_NAT_ATTR_SRC
:
1231 case OVS_NAT_ATTR_DST
:
1233 OVS_NLERR(log
, "Only one type of NAT may be specified");
1236 info
->nat
|= OVS_CT_NAT
;
1237 info
->nat
|= ((type
== OVS_NAT_ATTR_SRC
)
1238 ? OVS_CT_SRC_NAT
: OVS_CT_DST_NAT
);
1241 case OVS_NAT_ATTR_IP_MIN
:
1242 nla_memcpy(&info
->range
.min_addr
, a
,
1243 sizeof(info
->range
.min_addr
));
1244 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1247 case OVS_NAT_ATTR_IP_MAX
:
1249 nla_memcpy(&info
->range
.max_addr
, a
,
1250 sizeof(info
->range
.max_addr
));
1251 info
->range
.flags
|= NF_NAT_RANGE_MAP_IPS
;
1254 case OVS_NAT_ATTR_PROTO_MIN
:
1255 info
->range
.min_proto
.all
= htons(nla_get_u16(a
));
1256 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1259 case OVS_NAT_ATTR_PROTO_MAX
:
1260 have_proto_max
= true;
1261 info
->range
.max_proto
.all
= htons(nla_get_u16(a
));
1262 info
->range
.flags
|= NF_NAT_RANGE_PROTO_SPECIFIED
;
1265 case OVS_NAT_ATTR_PERSISTENT
:
1266 info
->range
.flags
|= NF_NAT_RANGE_PERSISTENT
;
1269 case OVS_NAT_ATTR_PROTO_HASH
:
1270 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM
;
1273 case OVS_NAT_ATTR_PROTO_RANDOM
:
1274 info
->range
.flags
|= NF_NAT_RANGE_PROTO_RANDOM_FULLY
;
1278 OVS_NLERR(log
, "Unknown nat attribute (%d)", type
);
1284 OVS_NLERR(log
, "NAT attribute has %d unknown bytes", rem
);
1288 /* Do not allow flags if no type is given. */
1289 if (info
->range
.flags
) {
1291 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1295 info
->nat
= OVS_CT_NAT
; /* NAT existing connections. */
1296 } else if (!info
->commit
) {
1298 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1302 /* Allow missing IP_MAX. */
1303 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
&& !have_ip_max
) {
1304 memcpy(&info
->range
.max_addr
, &info
->range
.min_addr
,
1305 sizeof(info
->range
.max_addr
));
1307 /* Allow missing PROTO_MAX. */
1308 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1310 info
->range
.max_proto
.all
= info
->range
.min_proto
.all
;
1316 static const struct ovs_ct_len_tbl ovs_ct_attr_lens
[OVS_CT_ATTR_MAX
+ 1] = {
1317 [OVS_CT_ATTR_COMMIT
] = { .minlen
= 0, .maxlen
= 0 },
1318 [OVS_CT_ATTR_FORCE_COMMIT
] = { .minlen
= 0, .maxlen
= 0 },
1319 [OVS_CT_ATTR_ZONE
] = { .minlen
= sizeof(u16
),
1320 .maxlen
= sizeof(u16
) },
1321 [OVS_CT_ATTR_MARK
] = { .minlen
= sizeof(struct md_mark
),
1322 .maxlen
= sizeof(struct md_mark
) },
1323 [OVS_CT_ATTR_LABELS
] = { .minlen
= sizeof(struct md_labels
),
1324 .maxlen
= sizeof(struct md_labels
) },
1325 [OVS_CT_ATTR_HELPER
] = { .minlen
= 1,
1326 .maxlen
= NF_CT_HELPER_NAME_LEN
},
1327 #ifdef CONFIG_NF_NAT_NEEDED
1328 /* NAT length is checked when parsing the nested attributes. */
1329 [OVS_CT_ATTR_NAT
] = { .minlen
= 0, .maxlen
= INT_MAX
},
1331 [OVS_CT_ATTR_EVENTMASK
] = { .minlen
= sizeof(u32
),
1332 .maxlen
= sizeof(u32
) },
1335 static int parse_ct(const struct nlattr
*attr
, struct ovs_conntrack_info
*info
,
1336 const char **helper
, bool log
)
1341 nla_for_each_nested(a
, attr
, rem
) {
1342 int type
= nla_type(a
);
1346 if (type
> OVS_CT_ATTR_MAX
) {
1348 "Unknown conntrack attr (type=%d, max=%d)",
1349 type
, OVS_CT_ATTR_MAX
);
1353 maxlen
= ovs_ct_attr_lens
[type
].maxlen
;
1354 minlen
= ovs_ct_attr_lens
[type
].minlen
;
1355 if (nla_len(a
) < minlen
|| nla_len(a
) > maxlen
) {
1357 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1358 type
, nla_len(a
), maxlen
);
1363 case OVS_CT_ATTR_FORCE_COMMIT
:
1366 case OVS_CT_ATTR_COMMIT
:
1367 info
->commit
= true;
1369 #ifdef CONFIG_NF_CONNTRACK_ZONES
1370 case OVS_CT_ATTR_ZONE
:
1371 info
->zone
.id
= nla_get_u16(a
);
1374 #ifdef CONFIG_NF_CONNTRACK_MARK
1375 case OVS_CT_ATTR_MARK
: {
1376 struct md_mark
*mark
= nla_data(a
);
1379 OVS_NLERR(log
, "ct_mark mask cannot be 0");
1386 #ifdef CONFIG_NF_CONNTRACK_LABELS
1387 case OVS_CT_ATTR_LABELS
: {
1388 struct md_labels
*labels
= nla_data(a
);
1390 if (!labels_nonzero(&labels
->mask
)) {
1391 OVS_NLERR(log
, "ct_labels mask cannot be 0");
1394 info
->labels
= *labels
;
1398 case OVS_CT_ATTR_HELPER
:
1399 *helper
= nla_data(a
);
1400 if (!memchr(*helper
, '\0', nla_len(a
))) {
1401 OVS_NLERR(log
, "Invalid conntrack helper");
1405 #ifdef CONFIG_NF_NAT_NEEDED
1406 case OVS_CT_ATTR_NAT
: {
1407 int err
= parse_nat(a
, info
, log
);
1414 case OVS_CT_ATTR_EVENTMASK
:
1415 info
->have_eventmask
= true;
1416 info
->eventmask
= nla_get_u32(a
);
1420 OVS_NLERR(log
, "Unknown conntrack attr (%d)",
1426 #ifdef CONFIG_NF_CONNTRACK_MARK
1427 if (!info
->commit
&& info
->mark
.mask
) {
1429 "Setting conntrack mark requires 'commit' flag.");
1433 #ifdef CONFIG_NF_CONNTRACK_LABELS
1434 if (!info
->commit
&& labels_nonzero(&info
->labels
.mask
)) {
1436 "Setting conntrack labels requires 'commit' flag.");
1441 OVS_NLERR(log
, "Conntrack attr has %d unknown bytes", rem
);
1448 bool ovs_ct_verify(struct net
*net
, enum ovs_key_attr attr
)
1450 if (attr
== OVS_KEY_ATTR_CT_STATE
)
1452 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1453 attr
== OVS_KEY_ATTR_CT_ZONE
)
1455 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) &&
1456 attr
== OVS_KEY_ATTR_CT_MARK
)
1458 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1459 attr
== OVS_KEY_ATTR_CT_LABELS
) {
1460 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1462 return ovs_net
->xt_label
;
1468 int ovs_ct_copy_action(struct net
*net
, const struct nlattr
*attr
,
1469 const struct sw_flow_key
*key
,
1470 struct sw_flow_actions
**sfa
, bool log
)
1472 struct ovs_conntrack_info ct_info
;
1473 const char *helper
= NULL
;
1477 family
= key_to_nfproto(key
);
1478 if (family
== NFPROTO_UNSPEC
) {
1479 OVS_NLERR(log
, "ct family unspecified");
1483 memset(&ct_info
, 0, sizeof(ct_info
));
1484 ct_info
.family
= family
;
1486 nf_ct_zone_init(&ct_info
.zone
, NF_CT_DEFAULT_ZONE_ID
,
1487 NF_CT_DEFAULT_ZONE_DIR
, 0);
1489 err
= parse_ct(attr
, &ct_info
, &helper
, log
);
1493 /* Set up template for tracking connections in specific zones. */
1494 ct_info
.ct
= nf_ct_tmpl_alloc(net
, &ct_info
.zone
, GFP_KERNEL
);
1496 OVS_NLERR(log
, "Failed to allocate conntrack template");
1500 __set_bit(IPS_CONFIRMED_BIT
, &ct_info
.ct
->status
);
1501 nf_conntrack_get(&ct_info
.ct
->ct_general
);
1504 err
= ovs_ct_add_helper(&ct_info
, helper
, key
, log
);
1509 err
= ovs_nla_add_action(sfa
, OVS_ACTION_ATTR_CT
, &ct_info
,
1510 sizeof(ct_info
), log
);
1516 __ovs_ct_free_action(&ct_info
);
1520 #ifdef CONFIG_NF_NAT_NEEDED
1521 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info
*info
,
1522 struct sk_buff
*skb
)
1524 struct nlattr
*start
;
1526 start
= nla_nest_start(skb
, OVS_CT_ATTR_NAT
);
1530 if (info
->nat
& OVS_CT_SRC_NAT
) {
1531 if (nla_put_flag(skb
, OVS_NAT_ATTR_SRC
))
1533 } else if (info
->nat
& OVS_CT_DST_NAT
) {
1534 if (nla_put_flag(skb
, OVS_NAT_ATTR_DST
))
1540 if (info
->range
.flags
& NF_NAT_RANGE_MAP_IPS
) {
1541 if (IS_ENABLED(CONFIG_NF_NAT_IPV4
) &&
1542 info
->family
== NFPROTO_IPV4
) {
1543 if (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1544 info
->range
.min_addr
.ip
) ||
1545 (info
->range
.max_addr
.ip
1546 != info
->range
.min_addr
.ip
&&
1547 (nla_put_in_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1548 info
->range
.max_addr
.ip
))))
1550 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6
) &&
1551 info
->family
== NFPROTO_IPV6
) {
1552 if (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MIN
,
1553 &info
->range
.min_addr
.in6
) ||
1554 (memcmp(&info
->range
.max_addr
.in6
,
1555 &info
->range
.min_addr
.in6
,
1556 sizeof(info
->range
.max_addr
.in6
)) &&
1557 (nla_put_in6_addr(skb
, OVS_NAT_ATTR_IP_MAX
,
1558 &info
->range
.max_addr
.in6
))))
1564 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_SPECIFIED
&&
1565 (nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MIN
,
1566 ntohs(info
->range
.min_proto
.all
)) ||
1567 (info
->range
.max_proto
.all
!= info
->range
.min_proto
.all
&&
1568 nla_put_u16(skb
, OVS_NAT_ATTR_PROTO_MAX
,
1569 ntohs(info
->range
.max_proto
.all
)))))
1572 if (info
->range
.flags
& NF_NAT_RANGE_PERSISTENT
&&
1573 nla_put_flag(skb
, OVS_NAT_ATTR_PERSISTENT
))
1575 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM
&&
1576 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_HASH
))
1578 if (info
->range
.flags
& NF_NAT_RANGE_PROTO_RANDOM_FULLY
&&
1579 nla_put_flag(skb
, OVS_NAT_ATTR_PROTO_RANDOM
))
1582 nla_nest_end(skb
, start
);
1588 int ovs_ct_action_to_attr(const struct ovs_conntrack_info
*ct_info
,
1589 struct sk_buff
*skb
)
1591 struct nlattr
*start
;
1593 start
= nla_nest_start(skb
, OVS_ACTION_ATTR_CT
);
1597 if (ct_info
->commit
&& nla_put_flag(skb
, ct_info
->force
1598 ? OVS_CT_ATTR_FORCE_COMMIT
1599 : OVS_CT_ATTR_COMMIT
))
1601 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES
) &&
1602 nla_put_u16(skb
, OVS_CT_ATTR_ZONE
, ct_info
->zone
.id
))
1604 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK
) && ct_info
->mark
.mask
&&
1605 nla_put(skb
, OVS_CT_ATTR_MARK
, sizeof(ct_info
->mark
),
1608 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS
) &&
1609 labels_nonzero(&ct_info
->labels
.mask
) &&
1610 nla_put(skb
, OVS_CT_ATTR_LABELS
, sizeof(ct_info
->labels
),
1613 if (ct_info
->helper
) {
1614 if (nla_put_string(skb
, OVS_CT_ATTR_HELPER
,
1615 ct_info
->helper
->name
))
1618 if (ct_info
->have_eventmask
&&
1619 nla_put_u32(skb
, OVS_CT_ATTR_EVENTMASK
, ct_info
->eventmask
))
1622 #ifdef CONFIG_NF_NAT_NEEDED
1623 if (ct_info
->nat
&& !ovs_ct_nat_to_attr(ct_info
, skb
))
1626 nla_nest_end(skb
, start
);
1631 void ovs_ct_free_action(const struct nlattr
*a
)
1633 struct ovs_conntrack_info
*ct_info
= nla_data(a
);
1635 __ovs_ct_free_action(ct_info
);
1638 static void __ovs_ct_free_action(struct ovs_conntrack_info
*ct_info
)
1640 if (ct_info
->helper
)
1641 nf_conntrack_helper_put(ct_info
->helper
);
1643 nf_ct_tmpl_free(ct_info
->ct
);
1646 void ovs_ct_init(struct net
*net
)
1648 unsigned int n_bits
= sizeof(struct ovs_key_ct_labels
) * BITS_PER_BYTE
;
1649 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1651 if (nf_connlabels_get(net
, n_bits
- 1)) {
1652 ovs_net
->xt_label
= false;
1653 OVS_NLERR(true, "Failed to set connlabel length");
1655 ovs_net
->xt_label
= true;
1659 void ovs_ct_exit(struct net
*net
)
1661 struct ovs_net
*ovs_net
= net_generic(net
, ovs_net_id
);
1663 if (ovs_net
->xt_label
)
1664 nf_connlabels_put(net
);