projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branches 'pm-core', 'pm-clk', 'pm-domains' and 'pm-pci'
[deliverable/linux.git] / net / openvswitch / conntrack.c
1 /*
2 * Copyright (c) 2015 Nicira, Inc.
3 *
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.
7 *
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.
12 */
13
14 #include <linux/module.h>
15 #include <linux/openvswitch.h>
16 #include <linux/tcp.h>
17 #include <linux/udp.h>
18 #include <linux/sctp.h>
19 #include <net/ip.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>
26
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>
31 #endif
32
33 #include "datapath.h"
34 #include "conntrack.h"
35 #include "flow.h"
36 #include "flow_netlink.h"
37
38 struct ovs_ct_len_tbl {
39 int maxlen;
40 int minlen;
41 };
42
43 /* Metadata mark for masked write to conntrack mark */
44 struct md_mark {
45 u32 value;
46 u32 mask;
47 };
48
49 /* Metadata label for masked write to conntrack label. */
50 struct md_labels {
51 struct ovs_key_ct_labels value;
52 struct ovs_key_ct_labels mask;
53 };
54
55 enum ovs_ct_nat {
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. */
59 };
60
61 /* Conntrack action context for execution. */
62 struct ovs_conntrack_info {
63 struct nf_conntrack_helper *helper;
64 struct nf_conntrack_zone zone;
65 struct nf_conn *ct;
66 u8 commit : 1;
67 u8 nat : 3; /* enum ovs_ct_nat */
68 u16 family;
69 struct md_mark mark;
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. */
73 #endif
74 };
75
76 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
77
78 static u16 key_to_nfproto(const struct sw_flow_key *key)
79 {
80 switch (ntohs(key->eth.type)) {
81 case ETH_P_IP:
82 return NFPROTO_IPV4;
83 case ETH_P_IPV6:
84 return NFPROTO_IPV6;
85 default:
86 return NFPROTO_UNSPEC;
87 }
88 }
89
90 /* Map SKB connection state into the values used by flow definition. */
91 static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
92 {
93 u8 ct_state = OVS_CS_F_TRACKED;
94
95 switch (ctinfo) {
96 case IP_CT_ESTABLISHED_REPLY:
97 case IP_CT_RELATED_REPLY:
98 ct_state |= OVS_CS_F_REPLY_DIR;
99 break;
100 default:
101 break;
102 }
103
104 switch (ctinfo) {
105 case IP_CT_ESTABLISHED:
106 case IP_CT_ESTABLISHED_REPLY:
107 ct_state |= OVS_CS_F_ESTABLISHED;
108 break;
109 case IP_CT_RELATED:
110 case IP_CT_RELATED_REPLY:
111 ct_state |= OVS_CS_F_RELATED;
112 break;
113 case IP_CT_NEW:
114 ct_state |= OVS_CS_F_NEW;
115 break;
116 default:
117 break;
118 }
119
120 return ct_state;
121 }
122
123 static u32 ovs_ct_get_mark(const struct nf_conn *ct)
124 {
125 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
126 return ct ? ct->mark : 0;
127 #else
128 return 0;
129 #endif
130 }
131
132 static void ovs_ct_get_labels(const struct nf_conn *ct,
133 struct ovs_key_ct_labels *labels)
134 {
135 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
136
137 if (cl) {
138 size_t len = cl->words * sizeof(long);
139
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);
145 } else {
146 memset(labels, 0, OVS_CT_LABELS_LEN);
147 }
148 }
149
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)
153 {
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);
158 }
159
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.
164 */
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,
168 bool keep_nat_flags)
169 {
170 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
171 enum ip_conntrack_info ctinfo;
172 struct nf_conn *ct;
173 u8 state = 0;
174
175 ct = nf_ct_get(skb, &ctinfo);
176 if (ct) {
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
182 * connection.
183 */
184 if (ct->master)
185 state |= OVS_CS_F_RELATED;
186 if (keep_nat_flags) {
187 state |= key->ct.state & OVS_CS_F_NAT_MASK;
188 } else {
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;
193 }
194 zone = nf_ct_zone(ct);
195 } else if (post_ct) {
196 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
197 if (info)
198 zone = &info->zone;
199 }
200 __ovs_ct_update_key(key, state, zone, ct);
201 }
202
203 /* This is called to initialize CT key fields possibly coming in from the local
204 * stack.
205 */
206 void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
207 {
208 ovs_ct_update_key(skb, NULL, key, false, false);
209 }
210
211 int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb)
212 {
213 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state))
214 return -EMSGSIZE;
215
216 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
217 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone))
218 return -EMSGSIZE;
219
220 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
221 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark))
222 return -EMSGSIZE;
223
224 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
225 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels),
226 &key->ct.labels))
227 return -EMSGSIZE;
228
229 return 0;
230 }
231
232 static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key,
233 u32 ct_mark, u32 mask)
234 {
235 #if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
236 enum ip_conntrack_info ctinfo;
237 struct nf_conn *ct;
238 u32 new_mark;
239
240 /* The connection could be invalid, in which case set_mark is no-op. */
241 ct = nf_ct_get(skb, &ctinfo);
242 if (!ct)
243 return 0;
244
245 new_mark = ct_mark | (ct->mark & ~(mask));
246 if (ct->mark != new_mark) {
247 ct->mark = new_mark;
248 nf_conntrack_event_cache(IPCT_MARK, ct);
249 key->ct.mark = new_mark;
250 }
251
252 return 0;
253 #else
254 return -ENOTSUPP;
255 #endif
256 }
257
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)
261 {
262 enum ip_conntrack_info ctinfo;
263 struct nf_conn_labels *cl;
264 struct nf_conn *ct;
265 int err;
266
267 /* The connection could be invalid, in which case set_label is no-op.*/
268 ct = nf_ct_get(skb, &ctinfo);
269 if (!ct)
270 return 0;
271
272 cl = nf_ct_labels_find(ct);
273 if (!cl) {
274 nf_ct_labels_ext_add(ct);
275 cl = nf_ct_labels_find(ct);
276 }
277 if (!cl || cl->words * sizeof(long) < OVS_CT_LABELS_LEN)
278 return -ENOSPC;
279
280 err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask,
281 OVS_CT_LABELS_LEN / sizeof(u32));
282 if (err)
283 return err;
284
285 ovs_ct_get_labels(ct, &key->ct.labels);
286 return 0;
287 }
288
289 /* 'skb' should already be pulled to nh_ofs. */
290 static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
291 {
292 const struct nf_conntrack_helper *helper;
293 const struct nf_conn_help *help;
294 enum ip_conntrack_info ctinfo;
295 unsigned int protoff;
296 struct nf_conn *ct;
297 int err;
298
299 ct = nf_ct_get(skb, &ctinfo);
300 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
301 return NF_ACCEPT;
302
303 help = nfct_help(ct);
304 if (!help)
305 return NF_ACCEPT;
306
307 helper = rcu_dereference(help->helper);
308 if (!helper)
309 return NF_ACCEPT;
310
311 switch (proto) {
312 case NFPROTO_IPV4:
313 protoff = ip_hdrlen(skb);
314 break;
315 case NFPROTO_IPV6: {
316 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
317 __be16 frag_off;
318 int ofs;
319
320 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
321 &frag_off);
322 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
323 pr_debug("proto header not found\n");
324 return NF_ACCEPT;
325 }
326 protoff = ofs;
327 break;
328 }
329 default:
330 WARN_ONCE(1, "helper invoked on non-IP family!");
331 return NF_DROP;
332 }
333
334 err = helper->help(skb, protoff, ct, ctinfo);
335 if (err != NF_ACCEPT)
336 return err;
337
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.
341 */
342 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
343 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
344 return NF_DROP;
345 return NF_ACCEPT;
346 }
347
348 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
349 * value if 'skb' is freed.
350 */
351 static int handle_fragments(struct net *net, struct sw_flow_key *key,
352 u16 zone, struct sk_buff *skb)
353 {
354 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
355 int err;
356
357 if (key->eth.type == htons(ETH_P_IP)) {
358 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
359
360 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
361 err = ip_defrag(net, skb, user);
362 if (err)
363 return err;
364
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;
369
370 skb_orphan(skb);
371 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
372 err = nf_ct_frag6_gather(net, skb, user);
373 if (err)
374 return err;
375
376 key->ip.proto = ipv6_hdr(skb)->nexthdr;
377 ovs_cb.mru = IP6CB(skb)->frag_max_size;
378 #endif
379 } else {
380 kfree_skb(skb);
381 return -EPFNOSUPPORT;
382 }
383
384 key->ip.frag = OVS_FRAG_TYPE_NONE;
385 skb_clear_hash(skb);
386 skb->ignore_df = 1;
387 *OVS_CB(skb) = ovs_cb;
388
389 return 0;
390 }
391
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)
395 {
396 struct nf_conntrack_tuple tuple;
397
398 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
399 return NULL;
400 return __nf_ct_expect_find(net, zone, &tuple);
401 }
402
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)
406 {
407 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
408
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;
416 return IP_CT_NEW;
417 }
418
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.
422 *
423 * Must be called with rcu_read_lock.
424 *
425 * On success, populates skb->nfct and skb->nfctinfo, and returns the
426 * connection. Returns NULL if there is no existing entry.
427 */
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)
431 {
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;
437 struct nf_conn *ct;
438 unsigned int dataoff;
439 u8 protonum;
440
441 l3proto = __nf_ct_l3proto_find(l3num);
442 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
443 &protonum) <= 0) {
444 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
445 return NULL;
446 }
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");
451 return NULL;
452 }
453
454 /* look for tuple match */
455 h = nf_conntrack_find_get(net, zone, &tuple);
456 if (!h)
457 return NULL; /* Not found. */
458
459 ct = nf_ct_tuplehash_to_ctrack(h);
460
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);
465 }
466 skb->nfct = &ct->ct_general;
467 skb->nfctinfo = ctinfo;
468 return ct;
469 }
470
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,
475 struct sk_buff *skb)
476 {
477 enum ip_conntrack_info ctinfo;
478 struct nf_conn *ct;
479
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.
485 */
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);
490 if (!ct)
491 return false;
492 if (!net_eq(net, read_pnet(&ct->ct_net)))
493 return false;
494 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
495 return false;
496 if (info->helper) {
497 struct nf_conn_help *help;
498
499 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
500 if (help && rcu_access_pointer(help->helper) != info->helper)
501 return false;
502 }
503
504 return true;
505 }
506
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.
511 */
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)
516 {
517 int hooknum, nh_off, err = NF_ACCEPT;
518
519 nh_off = skb_network_offset(skb);
520 skb_pull(skb, nh_off);
521
522 /* See HOOK2MANIP(). */
523 if (maniptype == NF_NAT_MANIP_SRC)
524 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
525 else
526 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
527
528 switch (ctinfo) {
529 case IP_CT_RELATED:
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,
535 hooknum))
536 err = NF_DROP;
537 goto push;
538 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
539 skb->protocol == htons(ETH_P_IPV6)) {
540 __be16 frag_off;
541 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
542 int hdrlen = ipv6_skip_exthdr(skb,
543 sizeof(struct ipv6hdr),
544 &nexthdr, &frag_off);
545
546 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
547 if (!nf_nat_icmpv6_reply_translation(skb, ct,
548 ctinfo,
549 hooknum,
550 hdrlen))
551 err = NF_DROP;
552 goto push;
553 }
554 }
555 /* Non-ICMP, fall thru to initialize if needed. */
556 case IP_CT_NEW:
557 /* Seen it before? This can happen for loopback, retrans,
558 * or local packets.
559 */
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
564 * mapping.
565 */
566 ? nf_nat_setup_info(ct, range, maniptype)
567 : nf_nat_alloc_null_binding(ct, hooknum);
568 if (err != NF_ACCEPT)
569 goto push;
570 }
571 break;
572
573 case IP_CT_ESTABLISHED:
574 case IP_CT_ESTABLISHED_REPLY:
575 break;
576
577 default:
578 err = NF_DROP;
579 goto push;
580 }
581
582 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
583 push:
584 skb_push(skb, nh_off);
585
586 return err;
587 }
588
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)
592 {
593 if (maniptype == NF_NAT_MANIP_SRC) {
594 __be16 src;
595
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));
602 else
603 return;
604
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;
611 else
612 return;
613
614 key->tp.src = src;
615 } else {
616 __be16 dst;
617
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));
624 else
625 return;
626
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;
633 else
634 return;
635
636 key->tp.dst = dst;
637 }
638 }
639
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)
645 {
646 enum nf_nat_manip_type maniptype;
647 int err;
648
649 if (nf_ct_is_untracked(ct)) {
650 /* A NAT action may only be performed on tracked packets. */
651 return NF_ACCEPT;
652 }
653
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. */
657
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
661 * when committing.
662 */
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.
671 */
672 maniptype = ct->status & IPS_SRC_NAT
673 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
674 else
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;
681 } else {
682 return NF_ACCEPT; /* Connection is not NATed. */
683 }
684 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
685
686 /* Mark NAT done if successful and update the flow key. */
687 if (err == NF_ACCEPT)
688 ovs_nat_update_key(key, skb, maniptype);
689
690 return err;
691 }
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)
697 {
698 return NF_ACCEPT;
699 }
700 #endif
701
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
704 * through conntrack.
705 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
706 * set to NULL and 0 will be returned.
707 */
708 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
709 const struct ovs_conntrack_info *info,
710 struct sk_buff *skb)
711 {
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
715 * different zone.
716 */
717 bool cached = skb_nfct_cached(net, key, info, skb);
718 enum ip_conntrack_info ctinfo;
719 struct nf_conn *ct;
720
721 if (!cached) {
722 struct nf_conn *tmpl = info->ct;
723 int err;
724
725 /* Associate skb with specified zone. */
726 if (tmpl) {
727 if (skb->nfct)
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;
732 }
733
734 /* Repeat if requested, see nf_iterate(). */
735 do {
736 err = nf_conntrack_in(net, info->family,
737 NF_INET_PRE_ROUTING, skb);
738 } while (err == NF_REPEAT);
739
740 if (err != NF_ACCEPT)
741 return -ENOENT;
742
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.
746 */
747 key->ct.state = 0;
748
749 /* Update the key, but keep the NAT flags. */
750 ovs_ct_update_key(skb, info, key, true, true);
751 }
752
753 ct = nf_ct_get(skb, &ctinfo);
754 if (ct) {
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.
760 *
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
763 * the key->ct.state.
764 */
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) {
768 return -EINVAL;
769 }
770
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.
775 */
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,
779 GFP_ATOMIC);
780 if (err)
781 return err;
782 }
783
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.
788 */
789 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
790 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
791 return -EINVAL;
792 }
793 }
794
795 return 0;
796 }
797
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,
801 struct sk_buff *skb)
802 {
803 struct nf_conntrack_expect *exp;
804
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.
811 */
812 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
813 if (exp) {
814 u8 state;
815
816 /* NOTE: New connections are NATted and Helped only when
817 * committed, so we are not calling into NAT here.
818 */
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);
821 } else {
822 struct nf_conn *ct;
823 int err;
824
825 err = __ovs_ct_lookup(net, key, info, skb);
826 if (err)
827 return err;
828
829 ct = (struct nf_conn *)skb->nfct;
830 if (ct)
831 nf_ct_deliver_cached_events(ct);
832 }
833
834 return 0;
835 }
836
837 /* Lookup connection and confirm if unconfirmed. */
838 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
839 const struct ovs_conntrack_info *info,
840 struct sk_buff *skb)
841 {
842 int err;
843
844 err = __ovs_ct_lookup(net, key, info, skb);
845 if (err)
846 return err;
847 /* This is a no-op if the connection has already been confirmed. */
848 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
849 return -EINVAL;
850
851 return 0;
852 }
853
854 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
855 {
856 size_t i;
857
858 for (i = 0; i < sizeof(*labels); i++)
859 if (labels->ct_labels[i])
860 return true;
861
862 return false;
863 }
864
865 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
866 * value if 'skb' is freed.
867 */
868 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
869 struct sw_flow_key *key,
870 const struct ovs_conntrack_info *info)
871 {
872 int nh_ofs;
873 int err;
874
875 /* The conntrack module expects to be working at L3. */
876 nh_ofs = skb_network_offset(skb);
877 skb_pull(skb, nh_ofs);
878
879 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
880 err = handle_fragments(net, key, info->zone.id, skb);
881 if (err)
882 return err;
883 }
884
885 if (info->commit)
886 err = ovs_ct_commit(net, key, info, skb);
887 else
888 err = ovs_ct_lookup(net, key, info, skb);
889 if (err)
890 goto err;
891
892 if (info->mark.mask) {
893 err = ovs_ct_set_mark(skb, key, info->mark.value,
894 info->mark.mask);
895 if (err)
896 goto err;
897 }
898 if (labels_nonzero(&info->labels.mask))
899 err = ovs_ct_set_labels(skb, key, &info->labels.value,
900 &info->labels.mask);
901 err:
902 skb_push(skb, nh_ofs);
903 if (err)
904 kfree_skb(skb);
905 return err;
906 }
907
908 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
909 const struct sw_flow_key *key, bool log)
910 {
911 struct nf_conntrack_helper *helper;
912 struct nf_conn_help *help;
913
914 helper = nf_conntrack_helper_try_module_get(name, info->family,
915 key->ip.proto);
916 if (!helper) {
917 OVS_NLERR(log, "Unknown helper \"%s\"", name);
918 return -EINVAL;
919 }
920
921 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
922 if (!help) {
923 module_put(helper->me);
924 return -ENOMEM;
925 }
926
927 rcu_assign_pointer(help->helper, helper);
928 info->helper = helper;
929 return 0;
930 }
931
932 #ifdef CONFIG_NF_NAT_NEEDED
933 static int parse_nat(const struct nlattr *attr,
934 struct ovs_conntrack_info *info, bool log)
935 {
936 struct nlattr *a;
937 int rem;
938 bool have_ip_max = false;
939 bool have_proto_max = false;
940 bool ip_vers = (info->family == NFPROTO_IPV6);
941
942 nla_for_each_nested(a, attr, rem) {
943 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
944 [OVS_NAT_ATTR_SRC] = {0, 0},
945 [OVS_NAT_ATTR_DST] = {0, 0},
946 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
947 sizeof(struct in6_addr)},
948 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
949 sizeof(struct in6_addr)},
950 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
951 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
952 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
953 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
954 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
955 };
956 int type = nla_type(a);
957
958 if (type > OVS_NAT_ATTR_MAX) {
959 OVS_NLERR(log,
960 "Unknown NAT attribute (type=%d, max=%d).\n",
961 type, OVS_NAT_ATTR_MAX);
962 return -EINVAL;
963 }
964
965 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
966 OVS_NLERR(log,
967 "NAT attribute type %d has unexpected length (%d != %d).\n",
968 type, nla_len(a),
969 ovs_nat_attr_lens[type][ip_vers]);
970 return -EINVAL;
971 }
972
973 switch (type) {
974 case OVS_NAT_ATTR_SRC:
975 case OVS_NAT_ATTR_DST:
976 if (info->nat) {
977 OVS_NLERR(log,
978 "Only one type of NAT may be specified.\n"
979 );
980 return -ERANGE;
981 }
982 info->nat |= OVS_CT_NAT;
983 info->nat |= ((type == OVS_NAT_ATTR_SRC)
984 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
985 break;
986
987 case OVS_NAT_ATTR_IP_MIN:
988 nla_memcpy(&info->range.min_addr, a,
989 sizeof(info->range.min_addr));
990 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
991 break;
992
993 case OVS_NAT_ATTR_IP_MAX:
994 have_ip_max = true;
995 nla_memcpy(&info->range.max_addr, a,
996 sizeof(info->range.max_addr));
997 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
998 break;
999
1000 case OVS_NAT_ATTR_PROTO_MIN:
1001 info->range.min_proto.all = htons(nla_get_u16(a));
1002 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1003 break;
1004
1005 case OVS_NAT_ATTR_PROTO_MAX:
1006 have_proto_max = true;
1007 info->range.max_proto.all = htons(nla_get_u16(a));
1008 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1009 break;
1010
1011 case OVS_NAT_ATTR_PERSISTENT:
1012 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1013 break;
1014
1015 case OVS_NAT_ATTR_PROTO_HASH:
1016 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1017 break;
1018
1019 case OVS_NAT_ATTR_PROTO_RANDOM:
1020 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1021 break;
1022
1023 default:
1024 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1025 return -EINVAL;
1026 }
1027 }
1028
1029 if (rem > 0) {
1030 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1031 return -EINVAL;
1032 }
1033 if (!info->nat) {
1034 /* Do not allow flags if no type is given. */
1035 if (info->range.flags) {
1036 OVS_NLERR(log,
1037 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1038 );
1039 return -EINVAL;
1040 }
1041 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1042 } else if (!info->commit) {
1043 OVS_NLERR(log,
1044 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1045 );
1046 return -EINVAL;
1047 }
1048 /* Allow missing IP_MAX. */
1049 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1050 memcpy(&info->range.max_addr, &info->range.min_addr,
1051 sizeof(info->range.max_addr));
1052 }
1053 /* Allow missing PROTO_MAX. */
1054 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1055 !have_proto_max) {
1056 info->range.max_proto.all = info->range.min_proto.all;
1057 }
1058 return 0;
1059 }
1060 #endif
1061
1062 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1063 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
1064 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1065 .maxlen = sizeof(u16) },
1066 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1067 .maxlen = sizeof(struct md_mark) },
1068 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1069 .maxlen = sizeof(struct md_labels) },
1070 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
1071 .maxlen = NF_CT_HELPER_NAME_LEN },
1072 #ifdef CONFIG_NF_NAT_NEEDED
1073 /* NAT length is checked when parsing the nested attributes. */
1074 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1075 #endif
1076 };
1077
1078 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1079 const char **helper, bool log)
1080 {
1081 struct nlattr *a;
1082 int rem;
1083
1084 nla_for_each_nested(a, attr, rem) {
1085 int type = nla_type(a);
1086 int maxlen = ovs_ct_attr_lens[type].maxlen;
1087 int minlen = ovs_ct_attr_lens[type].minlen;
1088
1089 if (type > OVS_CT_ATTR_MAX) {
1090 OVS_NLERR(log,
1091 "Unknown conntrack attr (type=%d, max=%d)",
1092 type, OVS_CT_ATTR_MAX);
1093 return -EINVAL;
1094 }
1095 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1096 OVS_NLERR(log,
1097 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1098 type, nla_len(a), maxlen);
1099 return -EINVAL;
1100 }
1101
1102 switch (type) {
1103 case OVS_CT_ATTR_COMMIT:
1104 info->commit = true;
1105 break;
1106 #ifdef CONFIG_NF_CONNTRACK_ZONES
1107 case OVS_CT_ATTR_ZONE:
1108 info->zone.id = nla_get_u16(a);
1109 break;
1110 #endif
1111 #ifdef CONFIG_NF_CONNTRACK_MARK
1112 case OVS_CT_ATTR_MARK: {
1113 struct md_mark *mark = nla_data(a);
1114
1115 if (!mark->mask) {
1116 OVS_NLERR(log, "ct_mark mask cannot be 0");
1117 return -EINVAL;
1118 }
1119 info->mark = *mark;
1120 break;
1121 }
1122 #endif
1123 #ifdef CONFIG_NF_CONNTRACK_LABELS
1124 case OVS_CT_ATTR_LABELS: {
1125 struct md_labels *labels = nla_data(a);
1126
1127 if (!labels_nonzero(&labels->mask)) {
1128 OVS_NLERR(log, "ct_labels mask cannot be 0");
1129 return -EINVAL;
1130 }
1131 info->labels = *labels;
1132 break;
1133 }
1134 #endif
1135 case OVS_CT_ATTR_HELPER:
1136 *helper = nla_data(a);
1137 if (!memchr(*helper, '\0', nla_len(a))) {
1138 OVS_NLERR(log, "Invalid conntrack helper");
1139 return -EINVAL;
1140 }
1141 break;
1142 #ifdef CONFIG_NF_NAT_NEEDED
1143 case OVS_CT_ATTR_NAT: {
1144 int err = parse_nat(a, info, log);
1145
1146 if (err)
1147 return err;
1148 break;
1149 }
1150 #endif
1151 default:
1152 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1153 type);
1154 return -EINVAL;
1155 }
1156 }
1157
1158 if (rem > 0) {
1159 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1160 return -EINVAL;
1161 }
1162
1163 return 0;
1164 }
1165
1166 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1167 {
1168 if (attr == OVS_KEY_ATTR_CT_STATE)
1169 return true;
1170 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1171 attr == OVS_KEY_ATTR_CT_ZONE)
1172 return true;
1173 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1174 attr == OVS_KEY_ATTR_CT_MARK)
1175 return true;
1176 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1177 attr == OVS_KEY_ATTR_CT_LABELS) {
1178 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1179
1180 return ovs_net->xt_label;
1181 }
1182
1183 return false;
1184 }
1185
1186 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1187 const struct sw_flow_key *key,
1188 struct sw_flow_actions **sfa, bool log)
1189 {
1190 struct ovs_conntrack_info ct_info;
1191 const char *helper = NULL;
1192 u16 family;
1193 int err;
1194
1195 family = key_to_nfproto(key);
1196 if (family == NFPROTO_UNSPEC) {
1197 OVS_NLERR(log, "ct family unspecified");
1198 return -EINVAL;
1199 }
1200
1201 memset(&ct_info, 0, sizeof(ct_info));
1202 ct_info.family = family;
1203
1204 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1205 NF_CT_DEFAULT_ZONE_DIR, 0);
1206
1207 err = parse_ct(attr, &ct_info, &helper, log);
1208 if (err)
1209 return err;
1210
1211 /* Set up template for tracking connections in specific zones. */
1212 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1213 if (!ct_info.ct) {
1214 OVS_NLERR(log, "Failed to allocate conntrack template");
1215 return -ENOMEM;
1216 }
1217
1218 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1219 nf_conntrack_get(&ct_info.ct->ct_general);
1220
1221 if (helper) {
1222 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1223 if (err)
1224 goto err_free_ct;
1225 }
1226
1227 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1228 sizeof(ct_info), log);
1229 if (err)
1230 goto err_free_ct;
1231
1232 return 0;
1233 err_free_ct:
1234 __ovs_ct_free_action(&ct_info);
1235 return err;
1236 }
1237
1238 #ifdef CONFIG_NF_NAT_NEEDED
1239 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1240 struct sk_buff *skb)
1241 {
1242 struct nlattr *start;
1243
1244 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1245 if (!start)
1246 return false;
1247
1248 if (info->nat & OVS_CT_SRC_NAT) {
1249 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1250 return false;
1251 } else if (info->nat & OVS_CT_DST_NAT) {
1252 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1253 return false;
1254 } else {
1255 goto out;
1256 }
1257
1258 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1259 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1260 info->family == NFPROTO_IPV4) {
1261 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1262 info->range.min_addr.ip) ||
1263 (info->range.max_addr.ip
1264 != info->range.min_addr.ip &&
1265 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1266 info->range.max_addr.ip))))
1267 return false;
1268 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1269 info->family == NFPROTO_IPV6) {
1270 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1271 &info->range.min_addr.in6) ||
1272 (memcmp(&info->range.max_addr.in6,
1273 &info->range.min_addr.in6,
1274 sizeof(info->range.max_addr.in6)) &&
1275 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1276 &info->range.max_addr.in6))))
1277 return false;
1278 } else {
1279 return false;
1280 }
1281 }
1282 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1283 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1284 ntohs(info->range.min_proto.all)) ||
1285 (info->range.max_proto.all != info->range.min_proto.all &&
1286 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1287 ntohs(info->range.max_proto.all)))))
1288 return false;
1289
1290 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1291 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1292 return false;
1293 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1294 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1295 return false;
1296 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1297 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1298 return false;
1299 out:
1300 nla_nest_end(skb, start);
1301
1302 return true;
1303 }
1304 #endif
1305
1306 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1307 struct sk_buff *skb)
1308 {
1309 struct nlattr *start;
1310
1311 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1312 if (!start)
1313 return -EMSGSIZE;
1314
1315 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
1316 return -EMSGSIZE;
1317 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1318 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1319 return -EMSGSIZE;
1320 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1321 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1322 &ct_info->mark))
1323 return -EMSGSIZE;
1324 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1325 labels_nonzero(&ct_info->labels.mask) &&
1326 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1327 &ct_info->labels))
1328 return -EMSGSIZE;
1329 if (ct_info->helper) {
1330 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1331 ct_info->helper->name))
1332 return -EMSGSIZE;
1333 }
1334 #ifdef CONFIG_NF_NAT_NEEDED
1335 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1336 return -EMSGSIZE;
1337 #endif
1338 nla_nest_end(skb, start);
1339
1340 return 0;
1341 }
1342
1343 void ovs_ct_free_action(const struct nlattr *a)
1344 {
1345 struct ovs_conntrack_info *ct_info = nla_data(a);
1346
1347 __ovs_ct_free_action(ct_info);
1348 }
1349
1350 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1351 {
1352 if (ct_info->helper)
1353 module_put(ct_info->helper->me);
1354 if (ct_info->ct)
1355 nf_ct_put(ct_info->ct);
1356 }
1357
1358 void ovs_ct_init(struct net *net)
1359 {
1360 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
1361 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1362
1363 if (nf_connlabels_get(net, n_bits - 1)) {
1364 ovs_net->xt_label = false;
1365 OVS_NLERR(true, "Failed to set connlabel length");
1366 } else {
1367 ovs_net->xt_label = true;
1368 }
1369 }
1370
1371 void ovs_ct_exit(struct net *net)
1372 {
1373 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1374
1375 if (ovs_net->xt_label)
1376 nf_connlabels_put(net);
1377 }
Linux kernel - Deliverables
This page took 0.064001 seconds and 5 git commands to generate.