?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
scsi: ufs: fix race between clock gating and devfreq scaling work
[linux/fpc-iii.git] / net / openvswitch / conntrack.c
blobf135814c34ad2762a920e412d147557723297ba1
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 = sizeof(cl->bits);
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 || sizeof(cl->bits) < 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 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
371 err = nf_ct_frag6_gather(net, skb, user);
372 if (err) {
373 if (err != -EINPROGRESS)
374 kfree_skb(skb);
375 return err;
376 }
377
378 key->ip.proto = ipv6_hdr(skb)->nexthdr;
379 ovs_cb.mru = IP6CB(skb)->frag_max_size;
380 #endif
381 } else {
382 kfree_skb(skb);
383 return -EPFNOSUPPORT;
384 }
385
386 key->ip.frag = OVS_FRAG_TYPE_NONE;
387 skb_clear_hash(skb);
388 skb->ignore_df = 1;
389 *OVS_CB(skb) = ovs_cb;
390
391 return 0;
392 }
393
394 static struct nf_conntrack_expect *
395 ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
396 u16 proto, const struct sk_buff *skb)
397 {
398 struct nf_conntrack_tuple tuple;
399 struct nf_conntrack_expect *exp;
400
401 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
402 return NULL;
403
404 exp = __nf_ct_expect_find(net, zone, &tuple);
405 if (exp) {
406 struct nf_conntrack_tuple_hash *h;
407
408 /* Delete existing conntrack entry, if it clashes with the
409 * expectation. This can happen since conntrack ALGs do not
410 * check for clashes between (new) expectations and existing
411 * conntrack entries. nf_conntrack_in() will check the
412 * expectations only if a conntrack entry can not be found,
413 * which can lead to OVS finding the expectation (here) in the
414 * init direction, but which will not be removed by the
415 * nf_conntrack_in() call, if a matching conntrack entry is
416 * found instead. In this case all init direction packets
417 * would be reported as new related packets, while reply
418 * direction packets would be reported as un-related
419 * established packets.
420 */
421 h = nf_conntrack_find_get(net, zone, &tuple);
422 if (h) {
423 struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
424
425 nf_ct_delete(ct, 0, 0);
426 nf_conntrack_put(&ct->ct_general);
427 }
428 }
429
430 return exp;
431 }
432
433 /* This replicates logic from nf_conntrack_core.c that is not exported. */
434 static enum ip_conntrack_info
435 ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
436 {
437 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
438
439 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
440 return IP_CT_ESTABLISHED_REPLY;
441 /* Once we've had two way comms, always ESTABLISHED. */
442 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
443 return IP_CT_ESTABLISHED;
444 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
445 return IP_CT_RELATED;
446 return IP_CT_NEW;
447 }
448
449 /* Find an existing connection which this packet belongs to without
450 * re-attributing statistics or modifying the connection state. This allows an
451 * skb->nfct lost due to an upcall to be recovered during actions execution.
452 *
453 * Must be called with rcu_read_lock.
454 *
455 * On success, populates skb->nfct and skb->nfctinfo, and returns the
456 * connection. Returns NULL if there is no existing entry.
457 */
458 static struct nf_conn *
459 ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
460 u8 l3num, struct sk_buff *skb)
461 {
462 struct nf_conntrack_l3proto *l3proto;
463 struct nf_conntrack_l4proto *l4proto;
464 struct nf_conntrack_tuple tuple;
465 struct nf_conntrack_tuple_hash *h;
466 struct nf_conn *ct;
467 unsigned int dataoff;
468 u8 protonum;
469
470 l3proto = __nf_ct_l3proto_find(l3num);
471 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
472 &protonum) <= 0) {
473 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
474 return NULL;
475 }
476 l4proto = __nf_ct_l4proto_find(l3num, protonum);
477 if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
478 protonum, net, &tuple, l3proto, l4proto)) {
479 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
480 return NULL;
481 }
482
483 /* look for tuple match */
484 h = nf_conntrack_find_get(net, zone, &tuple);
485 if (!h)
486 return NULL; /* Not found. */
487
488 ct = nf_ct_tuplehash_to_ctrack(h);
489
490 skb->nfct = &ct->ct_general;
491 skb->nfctinfo = ovs_ct_get_info(h);
492 return ct;
493 }
494
495 /* Determine whether skb->nfct is equal to the result of conntrack lookup. */
496 static bool skb_nfct_cached(struct net *net,
497 const struct sw_flow_key *key,
498 const struct ovs_conntrack_info *info,
499 struct sk_buff *skb)
500 {
501 enum ip_conntrack_info ctinfo;
502 struct nf_conn *ct;
503
504 ct = nf_ct_get(skb, &ctinfo);
505 /* If no ct, check if we have evidence that an existing conntrack entry
506 * might be found for this skb. This happens when we lose a skb->nfct
507 * due to an upcall. If the connection was not confirmed, it is not
508 * cached and needs to be run through conntrack again.
509 */
510 if (!ct && key->ct.state & OVS_CS_F_TRACKED &&
511 !(key->ct.state & OVS_CS_F_INVALID) &&
512 key->ct.zone == info->zone.id)
513 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb);
514 if (!ct)
515 return false;
516 if (!net_eq(net, read_pnet(&ct->ct_net)))
517 return false;
518 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
519 return false;
520 if (info->helper) {
521 struct nf_conn_help *help;
522
523 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
524 if (help && rcu_access_pointer(help->helper) != info->helper)
525 return false;
526 }
527
528 return true;
529 }
530
531 #ifdef CONFIG_NF_NAT_NEEDED
532 /* Modelled after nf_nat_ipv[46]_fn().
533 * range is only used for new, uninitialized NAT state.
534 * Returns either NF_ACCEPT or NF_DROP.
535 */
536 static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
537 enum ip_conntrack_info ctinfo,
538 const struct nf_nat_range *range,
539 enum nf_nat_manip_type maniptype)
540 {
541 int hooknum, nh_off, err = NF_ACCEPT;
542
543 nh_off = skb_network_offset(skb);
544 skb_pull_rcsum(skb, nh_off);
545
546 /* See HOOK2MANIP(). */
547 if (maniptype == NF_NAT_MANIP_SRC)
548 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
549 else
550 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
551
552 switch (ctinfo) {
553 case IP_CT_RELATED:
554 case IP_CT_RELATED_REPLY:
555 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
556 skb->protocol == htons(ETH_P_IP) &&
557 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
558 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
559 hooknum))
560 err = NF_DROP;
561 goto push;
562 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
563 skb->protocol == htons(ETH_P_IPV6)) {
564 __be16 frag_off;
565 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
566 int hdrlen = ipv6_skip_exthdr(skb,
567 sizeof(struct ipv6hdr),
568 &nexthdr, &frag_off);
569
570 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
571 if (!nf_nat_icmpv6_reply_translation(skb, ct,
572 ctinfo,
573 hooknum,
574 hdrlen))
575 err = NF_DROP;
576 goto push;
577 }
578 }
579 /* Non-ICMP, fall thru to initialize if needed. */
580 case IP_CT_NEW:
581 /* Seen it before? This can happen for loopback, retrans,
582 * or local packets.
583 */
584 if (!nf_nat_initialized(ct, maniptype)) {
585 /* Initialize according to the NAT action. */
586 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
587 /* Action is set up to establish a new
588 * mapping.
589 */
590 ? nf_nat_setup_info(ct, range, maniptype)
591 : nf_nat_alloc_null_binding(ct, hooknum);
592 if (err != NF_ACCEPT)
593 goto push;
594 }
595 break;
596
597 case IP_CT_ESTABLISHED:
598 case IP_CT_ESTABLISHED_REPLY:
599 break;
600
601 default:
602 err = NF_DROP;
603 goto push;
604 }
605
606 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
607 push:
608 skb_push(skb, nh_off);
609 skb_postpush_rcsum(skb, skb->data, nh_off);
610
611 return err;
612 }
613
614 static void ovs_nat_update_key(struct sw_flow_key *key,
615 const struct sk_buff *skb,
616 enum nf_nat_manip_type maniptype)
617 {
618 if (maniptype == NF_NAT_MANIP_SRC) {
619 __be16 src;
620
621 key->ct.state |= OVS_CS_F_SRC_NAT;
622 if (key->eth.type == htons(ETH_P_IP))
623 key->ipv4.addr.src = ip_hdr(skb)->saddr;
624 else if (key->eth.type == htons(ETH_P_IPV6))
625 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
626 sizeof(key->ipv6.addr.src));
627 else
628 return;
629
630 if (key->ip.proto == IPPROTO_UDP)
631 src = udp_hdr(skb)->source;
632 else if (key->ip.proto == IPPROTO_TCP)
633 src = tcp_hdr(skb)->source;
634 else if (key->ip.proto == IPPROTO_SCTP)
635 src = sctp_hdr(skb)->source;
636 else
637 return;
638
639 key->tp.src = src;
640 } else {
641 __be16 dst;
642
643 key->ct.state |= OVS_CS_F_DST_NAT;
644 if (key->eth.type == htons(ETH_P_IP))
645 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
646 else if (key->eth.type == htons(ETH_P_IPV6))
647 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
648 sizeof(key->ipv6.addr.dst));
649 else
650 return;
651
652 if (key->ip.proto == IPPROTO_UDP)
653 dst = udp_hdr(skb)->dest;
654 else if (key->ip.proto == IPPROTO_TCP)
655 dst = tcp_hdr(skb)->dest;
656 else if (key->ip.proto == IPPROTO_SCTP)
657 dst = sctp_hdr(skb)->dest;
658 else
659 return;
660
661 key->tp.dst = dst;
662 }
663 }
664
665 /* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
666 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
667 const struct ovs_conntrack_info *info,
668 struct sk_buff *skb, struct nf_conn *ct,
669 enum ip_conntrack_info ctinfo)
670 {
671 enum nf_nat_manip_type maniptype;
672 int err;
673
674 if (nf_ct_is_untracked(ct)) {
675 /* A NAT action may only be performed on tracked packets. */
676 return NF_ACCEPT;
677 }
678
679 /* Add NAT extension if not confirmed yet. */
680 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
681 return NF_ACCEPT; /* Can't NAT. */
682
683 /* Determine NAT type.
684 * Check if the NAT type can be deduced from the tracked connection.
685 * Make sure new expected connections (IP_CT_RELATED) are NATted only
686 * when committing.
687 */
688 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
689 ct->status & IPS_NAT_MASK &&
690 (ctinfo != IP_CT_RELATED || info->commit)) {
691 /* NAT an established or related connection like before. */
692 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
693 /* This is the REPLY direction for a connection
694 * for which NAT was applied in the forward
695 * direction. Do the reverse NAT.
696 */
697 maniptype = ct->status & IPS_SRC_NAT
698 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
699 else
700 maniptype = ct->status & IPS_SRC_NAT
701 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
702 } else if (info->nat & OVS_CT_SRC_NAT) {
703 maniptype = NF_NAT_MANIP_SRC;
704 } else if (info->nat & OVS_CT_DST_NAT) {
705 maniptype = NF_NAT_MANIP_DST;
706 } else {
707 return NF_ACCEPT; /* Connection is not NATed. */
708 }
709 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
710
711 /* Mark NAT done if successful and update the flow key. */
712 if (err == NF_ACCEPT)
713 ovs_nat_update_key(key, skb, maniptype);
714
715 return err;
716 }
717 #else /* !CONFIG_NF_NAT_NEEDED */
718 static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
719 const struct ovs_conntrack_info *info,
720 struct sk_buff *skb, struct nf_conn *ct,
721 enum ip_conntrack_info ctinfo)
722 {
723 return NF_ACCEPT;
724 }
725 #endif
726
727 /* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
728 * not done already. Update key with new CT state after passing the packet
729 * through conntrack.
730 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
731 * set to NULL and 0 will be returned.
732 */
733 static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
734 const struct ovs_conntrack_info *info,
735 struct sk_buff *skb)
736 {
737 /* If we are recirculating packets to match on conntrack fields and
738 * committing with a separate conntrack action, then we don't need to
739 * actually run the packet through conntrack twice unless it's for a
740 * different zone.
741 */
742 bool cached = skb_nfct_cached(net, key, info, skb);
743 enum ip_conntrack_info ctinfo;
744 struct nf_conn *ct;
745
746 if (!cached) {
747 struct nf_conn *tmpl = info->ct;
748 int err;
749
750 /* Associate skb with specified zone. */
751 if (tmpl) {
752 if (skb->nfct)
753 nf_conntrack_put(skb->nfct);
754 nf_conntrack_get(&tmpl->ct_general);
755 skb->nfct = &tmpl->ct_general;
756 skb->nfctinfo = IP_CT_NEW;
757 }
758
759 /* Repeat if requested, see nf_iterate(). */
760 do {
761 err = nf_conntrack_in(net, info->family,
762 NF_INET_PRE_ROUTING, skb);
763 } while (err == NF_REPEAT);
764
765 if (err != NF_ACCEPT)
766 return -ENOENT;
767
768 /* Clear CT state NAT flags to mark that we have not yet done
769 * NAT after the nf_conntrack_in() call. We can actually clear
770 * the whole state, as it will be re-initialized below.
771 */
772 key->ct.state = 0;
773
774 /* Update the key, but keep the NAT flags. */
775 ovs_ct_update_key(skb, info, key, true, true);
776 }
777
778 ct = nf_ct_get(skb, &ctinfo);
779 if (ct) {
780 /* Packets starting a new connection must be NATted before the
781 * helper, so that the helper knows about the NAT. We enforce
782 * this by delaying both NAT and helper calls for unconfirmed
783 * connections until the committing CT action. For later
784 * packets NAT and Helper may be called in either order.
785 *
786 * NAT will be done only if the CT action has NAT, and only
787 * once per packet (per zone), as guarded by the NAT bits in
788 * the key->ct.state.
789 */
790 if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) &&
791 (nf_ct_is_confirmed(ct) || info->commit) &&
792 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
793 return -EINVAL;
794 }
795
796 /* Userspace may decide to perform a ct lookup without a helper
797 * specified followed by a (recirculate and) commit with one.
798 * Therefore, for unconfirmed connections which we will commit,
799 * we need to attach the helper here.
800 */
801 if (!nf_ct_is_confirmed(ct) && info->commit &&
802 info->helper && !nfct_help(ct)) {
803 int err = __nf_ct_try_assign_helper(ct, info->ct,
804 GFP_ATOMIC);
805 if (err)
806 return err;
807 }
808
809 /* Call the helper only if:
810 * - nf_conntrack_in() was executed above ("!cached") for a
811 * confirmed connection, or
812 * - When committing an unconfirmed connection.
813 */
814 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
815 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
816 return -EINVAL;
817 }
818 }
819
820 return 0;
821 }
822
823 /* Lookup connection and read fields into key. */
824 static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
825 const struct ovs_conntrack_info *info,
826 struct sk_buff *skb)
827 {
828 struct nf_conntrack_expect *exp;
829
830 /* If we pass an expected packet through nf_conntrack_in() the
831 * expectation is typically removed, but the packet could still be
832 * lost in upcall processing. To prevent this from happening we
833 * perform an explicit expectation lookup. Expected connections are
834 * always new, and will be passed through conntrack only when they are
835 * committed, as it is OK to remove the expectation at that time.
836 */
837 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
838 if (exp) {
839 u8 state;
840
841 /* NOTE: New connections are NATted and Helped only when
842 * committed, so we are not calling into NAT here.
843 */
844 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
845 __ovs_ct_update_key(key, state, &info->zone, exp->master);
846 } else {
847 struct nf_conn *ct;
848 int err;
849
850 err = __ovs_ct_lookup(net, key, info, skb);
851 if (err)
852 return err;
853
854 ct = (struct nf_conn *)skb->nfct;
855 if (ct)
856 nf_ct_deliver_cached_events(ct);
857 }
858
859 return 0;
860 }
861
862 static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
863 {
864 size_t i;
865
866 for (i = 0; i < sizeof(*labels); i++)
867 if (labels->ct_labels[i])
868 return true;
869
870 return false;
871 }
872
873 /* Lookup connection and confirm if unconfirmed. */
874 static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
875 const struct ovs_conntrack_info *info,
876 struct sk_buff *skb)
877 {
878 int err;
879
880 err = __ovs_ct_lookup(net, key, info, skb);
881 if (err)
882 return err;
883
884 /* Apply changes before confirming the connection so that the initial
885 * conntrack NEW netlink event carries the values given in the CT
886 * action.
887 */
888 if (info->mark.mask) {
889 err = ovs_ct_set_mark(skb, key, info->mark.value,
890 info->mark.mask);
891 if (err)
892 return err;
893 }
894 if (labels_nonzero(&info->labels.mask)) {
895 err = ovs_ct_set_labels(skb, key, &info->labels.value,
896 &info->labels.mask);
897 if (err)
898 return err;
899 }
900 /* This will take care of sending queued events even if the connection
901 * is already confirmed.
902 */
903 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
904 return -EINVAL;
905
906 return 0;
907 }
908
909 /* Trim the skb to the length specified by the IP/IPv6 header,
910 * removing any trailing lower-layer padding. This prepares the skb
911 * for higher-layer processing that assumes skb->len excludes padding
912 * (such as nf_ip_checksum). The caller needs to pull the skb to the
913 * network header, and ensure ip_hdr/ipv6_hdr points to valid data.
914 */
915 static int ovs_skb_network_trim(struct sk_buff *skb)
916 {
917 unsigned int len;
918 int err;
919
920 switch (skb->protocol) {
921 case htons(ETH_P_IP):
922 len = ntohs(ip_hdr(skb)->tot_len);
923 break;
924 case htons(ETH_P_IPV6):
925 len = sizeof(struct ipv6hdr)
926 + ntohs(ipv6_hdr(skb)->payload_len);
927 break;
928 default:
929 len = skb->len;
930 }
931
932 err = pskb_trim_rcsum(skb, len);
933 if (err)
934 kfree_skb(skb);
935
936 return err;
937 }
938
939 /* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
940 * value if 'skb' is freed.
941 */
942 int ovs_ct_execute(struct net *net, struct sk_buff *skb,
943 struct sw_flow_key *key,
944 const struct ovs_conntrack_info *info)
945 {
946 int nh_ofs;
947 int err;
948
949 /* The conntrack module expects to be working at L3. */
950 nh_ofs = skb_network_offset(skb);
951 skb_pull_rcsum(skb, nh_ofs);
952
953 err = ovs_skb_network_trim(skb);
954 if (err)
955 return err;
956
957 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
958 err = handle_fragments(net, key, info->zone.id, skb);
959 if (err)
960 return err;
961 }
962
963 if (info->commit)
964 err = ovs_ct_commit(net, key, info, skb);
965 else
966 err = ovs_ct_lookup(net, key, info, skb);
967
968 skb_push(skb, nh_ofs);
969 skb_postpush_rcsum(skb, skb->data, nh_ofs);
970 if (err)
971 kfree_skb(skb);
972 return err;
973 }
974
975 static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
976 const struct sw_flow_key *key, bool log)
977 {
978 struct nf_conntrack_helper *helper;
979 struct nf_conn_help *help;
980
981 helper = nf_conntrack_helper_try_module_get(name, info->family,
982 key->ip.proto);
983 if (!helper) {
984 OVS_NLERR(log, "Unknown helper \"%s\"", name);
985 return -EINVAL;
986 }
987
988 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
989 if (!help) {
990 module_put(helper->me);
991 return -ENOMEM;
992 }
993
994 rcu_assign_pointer(help->helper, helper);
995 info->helper = helper;
996 return 0;
997 }
998
999 #ifdef CONFIG_NF_NAT_NEEDED
1000 static int parse_nat(const struct nlattr *attr,
1001 struct ovs_conntrack_info *info, bool log)
1002 {
1003 struct nlattr *a;
1004 int rem;
1005 bool have_ip_max = false;
1006 bool have_proto_max = false;
1007 bool ip_vers = (info->family == NFPROTO_IPV6);
1008
1009 nla_for_each_nested(a, attr, rem) {
1010 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
1011 [OVS_NAT_ATTR_SRC] = {0, 0},
1012 [OVS_NAT_ATTR_DST] = {0, 0},
1013 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
1014 sizeof(struct in6_addr)},
1015 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
1016 sizeof(struct in6_addr)},
1017 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
1018 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
1019 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
1020 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
1021 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
1022 };
1023 int type = nla_type(a);
1024
1025 if (type > OVS_NAT_ATTR_MAX) {
1026 OVS_NLERR(log,
1027 "Unknown NAT attribute (type=%d, max=%d).\n",
1028 type, OVS_NAT_ATTR_MAX);
1029 return -EINVAL;
1030 }
1031
1032 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
1033 OVS_NLERR(log,
1034 "NAT attribute type %d has unexpected length (%d != %d).\n",
1035 type, nla_len(a),
1036 ovs_nat_attr_lens[type][ip_vers]);
1037 return -EINVAL;
1038 }
1039
1040 switch (type) {
1041 case OVS_NAT_ATTR_SRC:
1042 case OVS_NAT_ATTR_DST:
1043 if (info->nat) {
1044 OVS_NLERR(log,
1045 "Only one type of NAT may be specified.\n"
1046 );
1047 return -ERANGE;
1048 }
1049 info->nat |= OVS_CT_NAT;
1050 info->nat |= ((type == OVS_NAT_ATTR_SRC)
1051 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
1052 break;
1053
1054 case OVS_NAT_ATTR_IP_MIN:
1055 nla_memcpy(&info->range.min_addr, a,
1056 sizeof(info->range.min_addr));
1057 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1058 break;
1059
1060 case OVS_NAT_ATTR_IP_MAX:
1061 have_ip_max = true;
1062 nla_memcpy(&info->range.max_addr, a,
1063 sizeof(info->range.max_addr));
1064 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1065 break;
1066
1067 case OVS_NAT_ATTR_PROTO_MIN:
1068 info->range.min_proto.all = htons(nla_get_u16(a));
1069 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1070 break;
1071
1072 case OVS_NAT_ATTR_PROTO_MAX:
1073 have_proto_max = true;
1074 info->range.max_proto.all = htons(nla_get_u16(a));
1075 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1076 break;
1077
1078 case OVS_NAT_ATTR_PERSISTENT:
1079 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1080 break;
1081
1082 case OVS_NAT_ATTR_PROTO_HASH:
1083 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1084 break;
1085
1086 case OVS_NAT_ATTR_PROTO_RANDOM:
1087 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1088 break;
1089
1090 default:
1091 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1092 return -EINVAL;
1093 }
1094 }
1095
1096 if (rem > 0) {
1097 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1098 return -EINVAL;
1099 }
1100 if (!info->nat) {
1101 /* Do not allow flags if no type is given. */
1102 if (info->range.flags) {
1103 OVS_NLERR(log,
1104 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1105 );
1106 return -EINVAL;
1107 }
1108 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1109 } else if (!info->commit) {
1110 OVS_NLERR(log,
1111 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1112 );
1113 return -EINVAL;
1114 }
1115 /* Allow missing IP_MAX. */
1116 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1117 memcpy(&info->range.max_addr, &info->range.min_addr,
1118 sizeof(info->range.max_addr));
1119 }
1120 /* Allow missing PROTO_MAX. */
1121 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1122 !have_proto_max) {
1123 info->range.max_proto.all = info->range.min_proto.all;
1124 }
1125 return 0;
1126 }
1127 #endif
1128
1129 static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
1130 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
1131 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1132 .maxlen = sizeof(u16) },
1133 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1134 .maxlen = sizeof(struct md_mark) },
1135 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1136 .maxlen = sizeof(struct md_labels) },
1137 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
1138 .maxlen = NF_CT_HELPER_NAME_LEN },
1139 #ifdef CONFIG_NF_NAT_NEEDED
1140 /* NAT length is checked when parsing the nested attributes. */
1141 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1142 #endif
1143 };
1144
1145 static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
1146 const char **helper, bool log)
1147 {
1148 struct nlattr *a;
1149 int rem;
1150
1151 nla_for_each_nested(a, attr, rem) {
1152 int type = nla_type(a);
1153 int maxlen;
1154 int minlen;
1155
1156 if (type > OVS_CT_ATTR_MAX) {
1157 OVS_NLERR(log,
1158 "Unknown conntrack attr (type=%d, max=%d)",
1159 type, OVS_CT_ATTR_MAX);
1160 return -EINVAL;
1161 }
1162
1163 maxlen = ovs_ct_attr_lens[type].maxlen;
1164 minlen = ovs_ct_attr_lens[type].minlen;
1165 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1166 OVS_NLERR(log,
1167 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1168 type, nla_len(a), maxlen);
1169 return -EINVAL;
1170 }
1171
1172 switch (type) {
1173 case OVS_CT_ATTR_COMMIT:
1174 info->commit = true;
1175 break;
1176 #ifdef CONFIG_NF_CONNTRACK_ZONES
1177 case OVS_CT_ATTR_ZONE:
1178 info->zone.id = nla_get_u16(a);
1179 break;
1180 #endif
1181 #ifdef CONFIG_NF_CONNTRACK_MARK
1182 case OVS_CT_ATTR_MARK: {
1183 struct md_mark *mark = nla_data(a);
1184
1185 if (!mark->mask) {
1186 OVS_NLERR(log, "ct_mark mask cannot be 0");
1187 return -EINVAL;
1188 }
1189 info->mark = *mark;
1190 break;
1191 }
1192 #endif
1193 #ifdef CONFIG_NF_CONNTRACK_LABELS
1194 case OVS_CT_ATTR_LABELS: {
1195 struct md_labels *labels = nla_data(a);
1196
1197 if (!labels_nonzero(&labels->mask)) {
1198 OVS_NLERR(log, "ct_labels mask cannot be 0");
1199 return -EINVAL;
1200 }
1201 info->labels = *labels;
1202 break;
1203 }
1204 #endif
1205 case OVS_CT_ATTR_HELPER:
1206 *helper = nla_data(a);
1207 if (!memchr(*helper, '\0', nla_len(a))) {
1208 OVS_NLERR(log, "Invalid conntrack helper");
1209 return -EINVAL;
1210 }
1211 break;
1212 #ifdef CONFIG_NF_NAT_NEEDED
1213 case OVS_CT_ATTR_NAT: {
1214 int err = parse_nat(a, info, log);
1215
1216 if (err)
1217 return err;
1218 break;
1219 }
1220 #endif
1221 default:
1222 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1223 type);
1224 return -EINVAL;
1225 }
1226 }
1227
1228 #ifdef CONFIG_NF_CONNTRACK_MARK
1229 if (!info->commit && info->mark.mask) {
1230 OVS_NLERR(log,
1231 "Setting conntrack mark requires 'commit' flag.");
1232 return -EINVAL;
1233 }
1234 #endif
1235 #ifdef CONFIG_NF_CONNTRACK_LABELS
1236 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1237 OVS_NLERR(log,
1238 "Setting conntrack labels requires 'commit' flag.");
1239 return -EINVAL;
1240 }
1241 #endif
1242 if (rem > 0) {
1243 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1244 return -EINVAL;
1245 }
1246
1247 return 0;
1248 }
1249
1250 bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
1251 {
1252 if (attr == OVS_KEY_ATTR_CT_STATE)
1253 return true;
1254 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1255 attr == OVS_KEY_ATTR_CT_ZONE)
1256 return true;
1257 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1258 attr == OVS_KEY_ATTR_CT_MARK)
1259 return true;
1260 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1261 attr == OVS_KEY_ATTR_CT_LABELS) {
1262 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1263
1264 return ovs_net->xt_label;
1265 }
1266
1267 return false;
1268 }
1269
1270 int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1271 const struct sw_flow_key *key,
1272 struct sw_flow_actions **sfa, bool log)
1273 {
1274 struct ovs_conntrack_info ct_info;
1275 const char *helper = NULL;
1276 u16 family;
1277 int err;
1278
1279 family = key_to_nfproto(key);
1280 if (family == NFPROTO_UNSPEC) {
1281 OVS_NLERR(log, "ct family unspecified");
1282 return -EINVAL;
1283 }
1284
1285 memset(&ct_info, 0, sizeof(ct_info));
1286 ct_info.family = family;
1287
1288 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1289 NF_CT_DEFAULT_ZONE_DIR, 0);
1290
1291 err = parse_ct(attr, &ct_info, &helper, log);
1292 if (err)
1293 return err;
1294
1295 /* Set up template for tracking connections in specific zones. */
1296 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1297 if (!ct_info.ct) {
1298 OVS_NLERR(log, "Failed to allocate conntrack template");
1299 return -ENOMEM;
1300 }
1301
1302 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1303 nf_conntrack_get(&ct_info.ct->ct_general);
1304
1305 if (helper) {
1306 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1307 if (err)
1308 goto err_free_ct;
1309 }
1310
1311 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1312 sizeof(ct_info), log);
1313 if (err)
1314 goto err_free_ct;
1315
1316 return 0;
1317 err_free_ct:
1318 __ovs_ct_free_action(&ct_info);
1319 return err;
1320 }
1321
1322 #ifdef CONFIG_NF_NAT_NEEDED
1323 static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1324 struct sk_buff *skb)
1325 {
1326 struct nlattr *start;
1327
1328 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1329 if (!start)
1330 return false;
1331
1332 if (info->nat & OVS_CT_SRC_NAT) {
1333 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1334 return false;
1335 } else if (info->nat & OVS_CT_DST_NAT) {
1336 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1337 return false;
1338 } else {
1339 goto out;
1340 }
1341
1342 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
1343 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1344 info->family == NFPROTO_IPV4) {
1345 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1346 info->range.min_addr.ip) ||
1347 (info->range.max_addr.ip
1348 != info->range.min_addr.ip &&
1349 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1350 info->range.max_addr.ip))))
1351 return false;
1352 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1353 info->family == NFPROTO_IPV6) {
1354 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1355 &info->range.min_addr.in6) ||
1356 (memcmp(&info->range.max_addr.in6,
1357 &info->range.min_addr.in6,
1358 sizeof(info->range.max_addr.in6)) &&
1359 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1360 &info->range.max_addr.in6))))
1361 return false;
1362 } else {
1363 return false;
1364 }
1365 }
1366 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1367 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1368 ntohs(info->range.min_proto.all)) ||
1369 (info->range.max_proto.all != info->range.min_proto.all &&
1370 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1371 ntohs(info->range.max_proto.all)))))
1372 return false;
1373
1374 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1375 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1376 return false;
1377 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1378 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1379 return false;
1380 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1381 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1382 return false;
1383 out:
1384 nla_nest_end(skb, start);
1385
1386 return true;
1387 }
1388 #endif
1389
1390 int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1391 struct sk_buff *skb)
1392 {
1393 struct nlattr *start;
1394
1395 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1396 if (!start)
1397 return -EMSGSIZE;
1398
1399 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
1400 return -EMSGSIZE;
1401 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1402 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1403 return -EMSGSIZE;
1404 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
1405 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1406 &ct_info->mark))
1407 return -EMSGSIZE;
1408 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
1409 labels_nonzero(&ct_info->labels.mask) &&
1410 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1411 &ct_info->labels))
1412 return -EMSGSIZE;
1413 if (ct_info->helper) {
1414 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1415 ct_info->helper->name))
1416 return -EMSGSIZE;
1417 }
1418 #ifdef CONFIG_NF_NAT_NEEDED
1419 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1420 return -EMSGSIZE;
1421 #endif
1422 nla_nest_end(skb, start);
1423
1424 return 0;
1425 }
1426
1427 void ovs_ct_free_action(const struct nlattr *a)
1428 {
1429 struct ovs_conntrack_info *ct_info = nla_data(a);
1430
1431 __ovs_ct_free_action(ct_info);
1432 }
1433
1434 static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1435 {
1436 if (ct_info->helper)
1437 module_put(ct_info->helper->me);
1438 if (ct_info->ct)
1439 nf_ct_tmpl_free(ct_info->ct);
1440 }
1441
1442 void ovs_ct_init(struct net *net)
1443 {
1444 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
1445 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1446
1447 if (nf_connlabels_get(net, n_bits - 1)) {
1448 ovs_net->xt_label = false;
1449 OVS_NLERR(true, "Failed to set connlabel length");
1450 } else {
1451 ovs_net->xt_label = true;
1452 }
1453 }
1454
1455 void ovs_ct_exit(struct net *net)
1456 {
1457 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1458
1459 if (ovs_net->xt_label)
1460 nf_connlabels_put(net);
1461 }
Linux with added FPC-III support