blob: 466393936db9492c99c41748bf0c1aa1a20a35f6 [file] [log] [blame]
Joe Stringer7f8a4362015-08-26 11:31:48 -07001/*
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>
Jarno Rajahalme05752522016-03-10 10:54:23 -080016#include <linux/tcp.h>
17#include <linux/udp.h>
18#include <linux/sctp.h>
Joe Stringer7f8a4362015-08-26 11:31:48 -070019#include <net/ip.h>
20#include <net/netfilter/nf_conntrack_core.h>
Joe Stringercae3a262015-08-26 11:31:53 -070021#include <net/netfilter/nf_conntrack_helper.h>
Joe Stringerc2ac6672015-08-26 11:31:52 -070022#include <net/netfilter/nf_conntrack_labels.h>
Jarno Rajahalme05752522016-03-10 10:54:23 -080023#include <net/netfilter/nf_conntrack_seqadj.h>
Joe Stringer7f8a4362015-08-26 11:31:48 -070024#include <net/netfilter/nf_conntrack_zones.h>
25#include <net/netfilter/ipv6/nf_defrag_ipv6.h>
26
Jarno Rajahalme05752522016-03-10 10:54:23 -080027#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
Joe Stringer7f8a4362015-08-26 11:31:48 -070033#include "datapath.h"
34#include "conntrack.h"
35#include "flow.h"
36#include "flow_netlink.h"
37
38struct ovs_ct_len_tbl {
Jarno Rajahalme05752522016-03-10 10:54:23 -080039 int maxlen;
40 int minlen;
Joe Stringer7f8a4362015-08-26 11:31:48 -070041};
42
Joe Stringer182e3042015-08-26 11:31:49 -070043/* Metadata mark for masked write to conntrack mark */
44struct md_mark {
45 u32 value;
46 u32 mask;
47};
48
Joe Stringerc2ac6672015-08-26 11:31:52 -070049/* Metadata label for masked write to conntrack label. */
Joe Stringer33db4122015-10-01 15:00:37 -070050struct md_labels {
51 struct ovs_key_ct_labels value;
52 struct ovs_key_ct_labels mask;
Joe Stringerc2ac6672015-08-26 11:31:52 -070053};
54
Jarno Rajahalme05752522016-03-10 10:54:23 -080055enum 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
Joe Stringer7f8a4362015-08-26 11:31:48 -070061/* Conntrack action context for execution. */
62struct ovs_conntrack_info {
Joe Stringercae3a262015-08-26 11:31:53 -070063 struct nf_conntrack_helper *helper;
Joe Stringer7f8a4362015-08-26 11:31:48 -070064 struct nf_conntrack_zone zone;
65 struct nf_conn *ct;
Joe Stringerab38a7b2015-10-06 11:00:01 -070066 u8 commit : 1;
Jarno Rajahalme05752522016-03-10 10:54:23 -080067 u8 nat : 3; /* enum ovs_ct_nat */
Joe Stringer7f8a4362015-08-26 11:31:48 -070068 u16 family;
Joe Stringer182e3042015-08-26 11:31:49 -070069 struct md_mark mark;
Joe Stringer33db4122015-10-01 15:00:37 -070070 struct md_labels labels;
Jarno Rajahalme05752522016-03-10 10:54:23 -080071#ifdef CONFIG_NF_NAT_NEEDED
72 struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */
73#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -070074};
75
Joe Stringer2f3ab9f2015-12-09 14:07:39 -080076static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
77
Joe Stringer7f8a4362015-08-26 11:31:48 -070078static 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. */
91static 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:
Joe Stringer7f8a4362015-08-26 11:31:48 -070098 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:
Joe Stringer7f8a4362015-08-26 11:31:48 -0700114 ct_state |= OVS_CS_F_NEW;
115 break;
116 default:
117 break;
118 }
119
120 return ct_state;
121}
122
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700123static 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
Joe Stringer33db4122015-10-01 15:00:37 -0700132static void ovs_ct_get_labels(const struct nf_conn *ct,
133 struct ovs_key_ct_labels *labels)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700134{
135 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
136
137 if (cl) {
Florian Westphal23014012016-07-21 12:51:16 +0200138 size_t len = sizeof(cl->bits);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700139
Joe Stringer33db4122015-10-01 15:00:37 -0700140 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);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700145 } else {
Joe Stringer33db4122015-10-01 15:00:37 -0700146 memset(labels, 0, OVS_CT_LABELS_LEN);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700147 }
148}
149
Joe Stringer7f8a4362015-08-26 11:31:48 -0700150static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
Joe Stringer182e3042015-08-26 11:31:49 -0700151 const struct nf_conntrack_zone *zone,
152 const struct nf_conn *ct)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700153{
154 key->ct.state = state;
155 key->ct.zone = zone->id;
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700156 key->ct.mark = ovs_ct_get_mark(ct);
Joe Stringer33db4122015-10-01 15:00:37 -0700157 ovs_ct_get_labels(ct, &key->ct.labels);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700158}
159
Jarno Rajahalme05752522016-03-10 10:54:23 -0800160/* 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.
Joe Stringer7f8a4362015-08-26 11:31:48 -0700164 */
165static void ovs_ct_update_key(const struct sk_buff *skb,
Joe Stringerd1109862015-12-09 14:07:40 -0800166 const struct ovs_conntrack_info *info,
Jarno Rajahalme05752522016-03-10 10:54:23 -0800167 struct sw_flow_key *key, bool post_ct,
168 bool keep_nat_flags)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700169{
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);
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800178 /* All unconfirmed entries are NEW connections. */
Joe Stringer4f0909e2015-10-19 19:18:59 -0700179 if (!nf_ct_is_confirmed(ct))
180 state |= OVS_CS_F_NEW;
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800181 /* OVS persists the related flag for the duration of the
182 * connection.
183 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700184 if (ct->master)
185 state |= OVS_CS_F_RELATED;
Jarno Rajahalme05752522016-03-10 10:54:23 -0800186 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 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700194 zone = nf_ct_zone(ct);
195 } else if (post_ct) {
196 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
Joe Stringerd1109862015-12-09 14:07:40 -0800197 if (info)
198 zone = &info->zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700199 }
Joe Stringer182e3042015-08-26 11:31:49 -0700200 __ovs_ct_update_key(key, state, zone, ct);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700201}
202
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800203/* This is called to initialize CT key fields possibly coming in from the local
204 * stack.
205 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700206void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
207{
Jarno Rajahalme05752522016-03-10 10:54:23 -0800208 ovs_ct_update_key(skb, NULL, key, false, false);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700209}
210
211int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb)
212{
Joe Stringerfbccce52015-10-06 11:00:00 -0700213 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state))
Joe Stringer7f8a4362015-08-26 11:31:48 -0700214 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
Joe Stringer182e3042015-08-26 11:31:49 -0700220 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
221 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark))
222 return -EMSGSIZE;
223
Valentin Rothberg9723e6a2015-08-28 10:39:56 +0200224 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringer33db4122015-10-01 15:00:37 -0700225 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels),
226 &key->ct.labels))
Joe Stringerc2ac6672015-08-26 11:31:52 -0700227 return -EMSGSIZE;
228
Joe Stringer182e3042015-08-26 11:31:49 -0700229 return 0;
230}
231
232static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key,
233 u32 ct_mark, u32 mask)
234{
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700235#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
Joe Stringer182e3042015-08-26 11:31:49 -0700236 enum ip_conntrack_info ctinfo;
237 struct nf_conn *ct;
238 u32 new_mark;
239
Joe Stringer182e3042015-08-26 11:31:49 -0700240 /* 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
Joe Stringer7f8a4362015-08-26 11:31:48 -0700252 return 0;
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700253#else
254 return -ENOTSUPP;
255#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -0700256}
257
Joe Stringer33db4122015-10-01 15:00:37 -0700258static 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)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700261{
262 enum ip_conntrack_info ctinfo;
263 struct nf_conn_labels *cl;
264 struct nf_conn *ct;
265 int err;
266
Joe Stringerc2ac6672015-08-26 11:31:52 -0700267 /* 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 }
Florian Westphal23014012016-07-21 12:51:16 +0200277 if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700278 return -ENOSPC;
279
Joe Stringer33db4122015-10-01 15:00:37 -0700280 err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask,
281 OVS_CT_LABELS_LEN / sizeof(u32));
Joe Stringerc2ac6672015-08-26 11:31:52 -0700282 if (err)
283 return err;
284
Joe Stringer33db4122015-10-01 15:00:37 -0700285 ovs_ct_get_labels(ct, &key->ct.labels);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700286 return 0;
287}
288
Joe Stringercae3a262015-08-26 11:31:53 -0700289/* 'skb' should already be pulled to nh_ofs. */
290static 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;
Jarno Rajahalme05752522016-03-10 10:54:23 -0800297 int err;
Joe Stringercae3a262015-08-26 11:31:53 -0700298
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;
Joe Stringercc570602015-09-14 11:14:50 -0700318 int ofs;
Joe Stringercae3a262015-08-26 11:31:53 -0700319
Joe Stringercc570602015-09-14 11:14:50 -0700320 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
321 &frag_off);
322 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
Joe Stringercae3a262015-08-26 11:31:53 -0700323 pr_debug("proto header not found\n");
324 return NF_ACCEPT;
325 }
Joe Stringercc570602015-09-14 11:14:50 -0700326 protoff = ofs;
Joe Stringercae3a262015-08-26 11:31:53 -0700327 break;
328 }
329 default:
330 WARN_ONCE(1, "helper invoked on non-IP family!");
331 return NF_DROP;
332 }
333
Jarno Rajahalme05752522016-03-10 10:54:23 -0800334 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;
Joe Stringercae3a262015-08-26 11:31:53 -0700346}
347
Joe Stringer74c16612015-10-25 20:21:48 -0700348/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
349 * value if 'skb' is freed.
350 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700351static 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);
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100355 int err;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700356
357 if (key->eth.type == htons(ETH_P_IP)) {
358 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700359
360 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500361 err = ip_defrag(net, skb, user);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700362 if (err)
363 return err;
364
365 ovs_cb.mru = IPCB(skb)->frag_max_size;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700366#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
Joe Stringer74c16612015-10-25 20:21:48 -0700367 } else if (key->eth.type == htons(ETH_P_IPV6)) {
Joe Stringer7f8a4362015-08-26 11:31:48 -0700368 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700369
370 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100371 err = nf_ct_frag6_gather(net, skb, user);
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800372 if (err) {
373 if (err != -EINPROGRESS)
374 kfree_skb(skb);
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100375 return err;
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800376 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700377
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100378 key->ip.proto = ipv6_hdr(skb)->nexthdr;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700379 ovs_cb.mru = IP6CB(skb)->frag_max_size;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700380#endif
381 } else {
Joe Stringer74c16612015-10-25 20:21:48 -0700382 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700383 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
394static struct nf_conntrack_expect *
395ovs_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;
Jarno Rajahalmed2e269c2017-04-14 14:26:38 -0700399 struct nf_conntrack_expect *exp;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700400
Eric W. Biedermana31f1ad2015-09-18 14:33:04 -0500401 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
Joe Stringer7f8a4362015-08-26 11:31:48 -0700402 return NULL;
Jarno Rajahalmed2e269c2017-04-14 14:26:38 -0700403
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;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700431}
432
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800433/* This replicates logic from nf_conntrack_core.c that is not exported. */
434static enum ip_conntrack_info
435ovs_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 */
458static struct nf_conn *
459ovs_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;
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800466 struct nf_conn *ct;
467 unsigned int dataoff;
468 u8 protonum;
469
470 l3proto = __nf_ct_l3proto_find(l3num);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800471 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);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800477 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
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800490 skb->nfct = &ct->ct_general;
Jarno Rajahalmebce91f82016-08-01 19:36:07 -0700491 skb->nfctinfo = ovs_ct_get_info(h);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800492 return ct;
493}
494
Joe Stringer7f8a4362015-08-26 11:31:48 -0700495/* Determine whether skb->nfct is equal to the result of conntrack lookup. */
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800496static 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)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700500{
501 enum ip_conntrack_info ctinfo;
502 struct nf_conn *ct;
503
504 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800505 /* 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);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700514 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;
Joe Stringercae3a262015-08-26 11:31:53 -0700520 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 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700527
528 return true;
529}
530
Jarno Rajahalme05752522016-03-10 10:54:23 -0800531#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 */
536static 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);
Lance Richardson18767ac2017-01-12 19:33:18 -0500544 skb_pull_rcsum(skb, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800545
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:
Arnd Bergmann99b72482016-03-18 14:33:45 +0100555 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
556 skb->protocol == htons(ETH_P_IP) &&
Jarno Rajahalme05752522016-03-10 10:54:23 -0800557 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;
Arnd Bergmann99b72482016-03-18 14:33:45 +0100562 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
563 skb->protocol == htons(ETH_P_IPV6)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800564 __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 }
Jarno Rajahalme05752522016-03-10 10:54:23 -0800578 }
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);
607push:
608 skb_push(skb, nh_off);
Lance Richardson18767ac2017-01-12 19:33:18 -0500609 skb_postpush_rcsum(skb, skb->data, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800610
611 return err;
612}
613
614static 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. */
666static 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.
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700685 * Make sure new expected connections (IP_CT_RELATED) are NATted only
686 * when committing.
Jarno Rajahalme05752522016-03-10 10:54:23 -0800687 */
688 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
689 ct->status & IPS_NAT_MASK &&
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700690 (ctinfo != IP_CT_RELATED || info->commit)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800691 /* 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 */
718static 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
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800727/* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800728 * not done already. Update key with new CT state after passing the packet
729 * through conntrack.
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800730 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
731 * set to NULL and 0 will be returned.
732 */
Joe Stringer4f0909e2015-10-19 19:18:59 -0700733static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
Joe Stringer7f8a4362015-08-26 11:31:48 -0700734 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 */
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800742 bool cached = skb_nfct_cached(net, key, info, skb);
743 enum ip_conntrack_info ctinfo;
744 struct nf_conn *ct;
745
746 if (!cached) {
Joe Stringer7f8a4362015-08-26 11:31:48 -0700747 struct nf_conn *tmpl = info->ct;
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800748 int err;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700749
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
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800759 /* 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)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700766 return -ENOENT;
Joe Stringercae3a262015-08-26 11:31:53 -0700767
Jarno Rajahalme05752522016-03-10 10:54:23 -0800768 /* 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);
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800776 }
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800777
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800778 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800779 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
Joe Stringer16ec3d42016-05-11 10:29:26 -0700796 /* 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
Jarno Rajahalme05752522016-03-10 10:54:23 -0800809 /* 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 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700818 }
819
820 return 0;
821}
822
823/* Lookup connection and read fields into key. */
824static 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
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800830 /* 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 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700837 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
838 if (exp) {
839 u8 state;
840
Jarno Rajahalme05752522016-03-10 10:54:23 -0800841 /* NOTE: New connections are NATted and Helped only when
842 * committed, so we are not calling into NAT here.
843 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700844 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
Joe Stringer182e3042015-08-26 11:31:49 -0700845 __ovs_ct_update_key(key, state, &info->zone, exp->master);
Samuel Gauthierd913d3a2016-06-28 17:22:26 +0200846 } 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 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700858
859 return 0;
860}
861
Joe Stringer33db4122015-10-01 15:00:37 -0700862static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700863{
864 size_t i;
865
Joe Stringer33db4122015-10-01 15:00:37 -0700866 for (i = 0; i < sizeof(*labels); i++)
867 if (labels->ct_labels[i])
Joe Stringerc2ac6672015-08-26 11:31:52 -0700868 return true;
869
870 return false;
871}
872
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700873/* Lookup connection and confirm if unconfirmed. */
874static 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
Joe Stringer74c16612015-10-25 20:21:48 -0700909/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
910 * value if 'skb' is freed.
911 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700912int ovs_ct_execute(struct net *net, struct sk_buff *skb,
913 struct sw_flow_key *key,
914 const struct ovs_conntrack_info *info)
915{
916 int nh_ofs;
917 int err;
918
919 /* The conntrack module expects to be working at L3. */
920 nh_ofs = skb_network_offset(skb);
Lance Richardson18767ac2017-01-12 19:33:18 -0500921 skb_pull_rcsum(skb, nh_ofs);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700922
923 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
924 err = handle_fragments(net, key, info->zone.id, skb);
925 if (err)
926 return err;
927 }
928
Joe Stringerab38a7b2015-10-06 11:00:01 -0700929 if (info->commit)
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700930 err = ovs_ct_commit(net, key, info, skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700931 else
932 err = ovs_ct_lookup(net, key, info, skb);
933
934 skb_push(skb, nh_ofs);
Lance Richardson18767ac2017-01-12 19:33:18 -0500935 skb_postpush_rcsum(skb, skb->data, nh_ofs);
Joe Stringer74c16612015-10-25 20:21:48 -0700936 if (err)
937 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700938 return err;
939}
940
Joe Stringercae3a262015-08-26 11:31:53 -0700941static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
942 const struct sw_flow_key *key, bool log)
943{
944 struct nf_conntrack_helper *helper;
945 struct nf_conn_help *help;
946
947 helper = nf_conntrack_helper_try_module_get(name, info->family,
948 key->ip.proto);
949 if (!helper) {
950 OVS_NLERR(log, "Unknown helper \"%s\"", name);
951 return -EINVAL;
952 }
953
954 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
955 if (!help) {
956 module_put(helper->me);
957 return -ENOMEM;
958 }
959
960 rcu_assign_pointer(help->helper, helper);
961 info->helper = helper;
962 return 0;
963}
964
Jarno Rajahalme05752522016-03-10 10:54:23 -0800965#ifdef CONFIG_NF_NAT_NEEDED
966static int parse_nat(const struct nlattr *attr,
967 struct ovs_conntrack_info *info, bool log)
968{
969 struct nlattr *a;
970 int rem;
971 bool have_ip_max = false;
972 bool have_proto_max = false;
973 bool ip_vers = (info->family == NFPROTO_IPV6);
974
975 nla_for_each_nested(a, attr, rem) {
976 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
977 [OVS_NAT_ATTR_SRC] = {0, 0},
978 [OVS_NAT_ATTR_DST] = {0, 0},
979 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
980 sizeof(struct in6_addr)},
981 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
982 sizeof(struct in6_addr)},
983 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
984 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
985 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
986 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
987 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
988 };
989 int type = nla_type(a);
990
991 if (type > OVS_NAT_ATTR_MAX) {
992 OVS_NLERR(log,
993 "Unknown NAT attribute (type=%d, max=%d).\n",
994 type, OVS_NAT_ATTR_MAX);
995 return -EINVAL;
996 }
997
998 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
999 OVS_NLERR(log,
1000 "NAT attribute type %d has unexpected length (%d != %d).\n",
1001 type, nla_len(a),
1002 ovs_nat_attr_lens[type][ip_vers]);
1003 return -EINVAL;
1004 }
1005
1006 switch (type) {
1007 case OVS_NAT_ATTR_SRC:
1008 case OVS_NAT_ATTR_DST:
1009 if (info->nat) {
1010 OVS_NLERR(log,
1011 "Only one type of NAT may be specified.\n"
1012 );
1013 return -ERANGE;
1014 }
1015 info->nat |= OVS_CT_NAT;
1016 info->nat |= ((type == OVS_NAT_ATTR_SRC)
1017 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
1018 break;
1019
1020 case OVS_NAT_ATTR_IP_MIN:
Haishuang Yanac71b462016-03-28 18:08:59 +08001021 nla_memcpy(&info->range.min_addr, a,
1022 sizeof(info->range.min_addr));
Jarno Rajahalme05752522016-03-10 10:54:23 -08001023 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1024 break;
1025
1026 case OVS_NAT_ATTR_IP_MAX:
1027 have_ip_max = true;
1028 nla_memcpy(&info->range.max_addr, a,
1029 sizeof(info->range.max_addr));
1030 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1031 break;
1032
1033 case OVS_NAT_ATTR_PROTO_MIN:
1034 info->range.min_proto.all = htons(nla_get_u16(a));
1035 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1036 break;
1037
1038 case OVS_NAT_ATTR_PROTO_MAX:
1039 have_proto_max = true;
1040 info->range.max_proto.all = htons(nla_get_u16(a));
1041 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1042 break;
1043
1044 case OVS_NAT_ATTR_PERSISTENT:
1045 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1046 break;
1047
1048 case OVS_NAT_ATTR_PROTO_HASH:
1049 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1050 break;
1051
1052 case OVS_NAT_ATTR_PROTO_RANDOM:
1053 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1054 break;
1055
1056 default:
1057 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1058 return -EINVAL;
1059 }
1060 }
1061
1062 if (rem > 0) {
1063 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1064 return -EINVAL;
1065 }
1066 if (!info->nat) {
1067 /* Do not allow flags if no type is given. */
1068 if (info->range.flags) {
1069 OVS_NLERR(log,
1070 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1071 );
1072 return -EINVAL;
1073 }
1074 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1075 } else if (!info->commit) {
1076 OVS_NLERR(log,
1077 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1078 );
1079 return -EINVAL;
1080 }
1081 /* Allow missing IP_MAX. */
1082 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1083 memcpy(&info->range.max_addr, &info->range.min_addr,
1084 sizeof(info->range.max_addr));
1085 }
1086 /* Allow missing PROTO_MAX. */
1087 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1088 !have_proto_max) {
1089 info->range.max_proto.all = info->range.min_proto.all;
1090 }
1091 return 0;
1092}
1093#endif
1094
Joe Stringer7f8a4362015-08-26 11:31:48 -07001095static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001096 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
Joe Stringer7f8a4362015-08-26 11:31:48 -07001097 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1098 .maxlen = sizeof(u16) },
Joe Stringer182e3042015-08-26 11:31:49 -07001099 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1100 .maxlen = sizeof(struct md_mark) },
Joe Stringer33db4122015-10-01 15:00:37 -07001101 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1102 .maxlen = sizeof(struct md_labels) },
Joe Stringercae3a262015-08-26 11:31:53 -07001103 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
Jarno Rajahalme05752522016-03-10 10:54:23 -08001104 .maxlen = NF_CT_HELPER_NAME_LEN },
1105#ifdef CONFIG_NF_NAT_NEEDED
1106 /* NAT length is checked when parsing the nested attributes. */
1107 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1108#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001109};
1110
1111static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
Joe Stringercae3a262015-08-26 11:31:53 -07001112 const char **helper, bool log)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001113{
1114 struct nlattr *a;
1115 int rem;
1116
1117 nla_for_each_nested(a, attr, rem) {
1118 int type = nla_type(a);
Liping Zhangd53ff382017-07-23 17:52:23 +08001119 int maxlen;
1120 int minlen;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001121
1122 if (type > OVS_CT_ATTR_MAX) {
1123 OVS_NLERR(log,
1124 "Unknown conntrack attr (type=%d, max=%d)",
1125 type, OVS_CT_ATTR_MAX);
1126 return -EINVAL;
1127 }
Liping Zhangd53ff382017-07-23 17:52:23 +08001128
1129 maxlen = ovs_ct_attr_lens[type].maxlen;
1130 minlen = ovs_ct_attr_lens[type].minlen;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001131 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1132 OVS_NLERR(log,
1133 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1134 type, nla_len(a), maxlen);
1135 return -EINVAL;
1136 }
1137
1138 switch (type) {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001139 case OVS_CT_ATTR_COMMIT:
1140 info->commit = true;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001141 break;
1142#ifdef CONFIG_NF_CONNTRACK_ZONES
1143 case OVS_CT_ATTR_ZONE:
1144 info->zone.id = nla_get_u16(a);
1145 break;
1146#endif
Joe Stringer182e3042015-08-26 11:31:49 -07001147#ifdef CONFIG_NF_CONNTRACK_MARK
1148 case OVS_CT_ATTR_MARK: {
1149 struct md_mark *mark = nla_data(a);
1150
Joe Stringere754ec62015-10-19 19:19:00 -07001151 if (!mark->mask) {
1152 OVS_NLERR(log, "ct_mark mask cannot be 0");
1153 return -EINVAL;
1154 }
Joe Stringer182e3042015-08-26 11:31:49 -07001155 info->mark = *mark;
1156 break;
1157 }
1158#endif
Joe Stringerc2ac6672015-08-26 11:31:52 -07001159#ifdef CONFIG_NF_CONNTRACK_LABELS
Joe Stringer33db4122015-10-01 15:00:37 -07001160 case OVS_CT_ATTR_LABELS: {
1161 struct md_labels *labels = nla_data(a);
Joe Stringerc2ac6672015-08-26 11:31:52 -07001162
Joe Stringere754ec62015-10-19 19:19:00 -07001163 if (!labels_nonzero(&labels->mask)) {
1164 OVS_NLERR(log, "ct_labels mask cannot be 0");
1165 return -EINVAL;
1166 }
Joe Stringer33db4122015-10-01 15:00:37 -07001167 info->labels = *labels;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001168 break;
1169 }
1170#endif
Joe Stringercae3a262015-08-26 11:31:53 -07001171 case OVS_CT_ATTR_HELPER:
1172 *helper = nla_data(a);
1173 if (!memchr(*helper, '\0', nla_len(a))) {
1174 OVS_NLERR(log, "Invalid conntrack helper");
1175 return -EINVAL;
1176 }
1177 break;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001178#ifdef CONFIG_NF_NAT_NEEDED
1179 case OVS_CT_ATTR_NAT: {
1180 int err = parse_nat(a, info, log);
1181
1182 if (err)
1183 return err;
1184 break;
1185 }
1186#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001187 default:
1188 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1189 type);
1190 return -EINVAL;
1191 }
1192 }
1193
Jarno Rajahalme7d904c72016-06-21 14:59:38 -07001194#ifdef CONFIG_NF_CONNTRACK_MARK
1195 if (!info->commit && info->mark.mask) {
1196 OVS_NLERR(log,
1197 "Setting conntrack mark requires 'commit' flag.");
1198 return -EINVAL;
1199 }
1200#endif
1201#ifdef CONFIG_NF_CONNTRACK_LABELS
1202 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1203 OVS_NLERR(log,
1204 "Setting conntrack labels requires 'commit' flag.");
1205 return -EINVAL;
1206 }
1207#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001208 if (rem > 0) {
1209 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1210 return -EINVAL;
1211 }
1212
1213 return 0;
1214}
1215
Joe Stringerc2ac6672015-08-26 11:31:52 -07001216bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001217{
1218 if (attr == OVS_KEY_ATTR_CT_STATE)
1219 return true;
1220 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1221 attr == OVS_KEY_ATTR_CT_ZONE)
1222 return true;
Joe Stringer182e3042015-08-26 11:31:49 -07001223 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1224 attr == OVS_KEY_ATTR_CT_MARK)
1225 return true;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001226 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001227 attr == OVS_KEY_ATTR_CT_LABELS) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001228 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1229
1230 return ovs_net->xt_label;
1231 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001232
1233 return false;
1234}
1235
1236int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1237 const struct sw_flow_key *key,
1238 struct sw_flow_actions **sfa, bool log)
1239{
1240 struct ovs_conntrack_info ct_info;
Joe Stringercae3a262015-08-26 11:31:53 -07001241 const char *helper = NULL;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001242 u16 family;
1243 int err;
1244
1245 family = key_to_nfproto(key);
1246 if (family == NFPROTO_UNSPEC) {
1247 OVS_NLERR(log, "ct family unspecified");
1248 return -EINVAL;
1249 }
1250
1251 memset(&ct_info, 0, sizeof(ct_info));
1252 ct_info.family = family;
1253
1254 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1255 NF_CT_DEFAULT_ZONE_DIR, 0);
1256
Joe Stringercae3a262015-08-26 11:31:53 -07001257 err = parse_ct(attr, &ct_info, &helper, log);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001258 if (err)
1259 return err;
1260
1261 /* Set up template for tracking connections in specific zones. */
1262 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1263 if (!ct_info.ct) {
1264 OVS_NLERR(log, "Failed to allocate conntrack template");
1265 return -ENOMEM;
1266 }
Joe Stringer90c7afc962015-12-23 14:39:27 -08001267
1268 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1269 nf_conntrack_get(&ct_info.ct->ct_general);
1270
Joe Stringercae3a262015-08-26 11:31:53 -07001271 if (helper) {
1272 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1273 if (err)
1274 goto err_free_ct;
1275 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001276
1277 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1278 sizeof(ct_info), log);
1279 if (err)
1280 goto err_free_ct;
1281
Joe Stringer7f8a4362015-08-26 11:31:48 -07001282 return 0;
1283err_free_ct:
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001284 __ovs_ct_free_action(&ct_info);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001285 return err;
1286}
1287
Jarno Rajahalme05752522016-03-10 10:54:23 -08001288#ifdef CONFIG_NF_NAT_NEEDED
1289static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1290 struct sk_buff *skb)
1291{
1292 struct nlattr *start;
1293
1294 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1295 if (!start)
1296 return false;
1297
1298 if (info->nat & OVS_CT_SRC_NAT) {
1299 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1300 return false;
1301 } else if (info->nat & OVS_CT_DST_NAT) {
1302 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1303 return false;
1304 } else {
1305 goto out;
1306 }
1307
1308 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
Arnd Bergmann99b72482016-03-18 14:33:45 +01001309 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1310 info->family == NFPROTO_IPV4) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001311 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1312 info->range.min_addr.ip) ||
1313 (info->range.max_addr.ip
1314 != info->range.min_addr.ip &&
1315 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1316 info->range.max_addr.ip))))
1317 return false;
Arnd Bergmann99b72482016-03-18 14:33:45 +01001318 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1319 info->family == NFPROTO_IPV6) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001320 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1321 &info->range.min_addr.in6) ||
1322 (memcmp(&info->range.max_addr.in6,
1323 &info->range.min_addr.in6,
1324 sizeof(info->range.max_addr.in6)) &&
1325 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1326 &info->range.max_addr.in6))))
1327 return false;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001328 } else {
1329 return false;
1330 }
1331 }
1332 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1333 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1334 ntohs(info->range.min_proto.all)) ||
1335 (info->range.max_proto.all != info->range.min_proto.all &&
1336 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1337 ntohs(info->range.max_proto.all)))))
1338 return false;
1339
1340 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1341 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1342 return false;
1343 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1344 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1345 return false;
1346 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1347 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1348 return false;
1349out:
1350 nla_nest_end(skb, start);
1351
1352 return true;
1353}
1354#endif
1355
Joe Stringer7f8a4362015-08-26 11:31:48 -07001356int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1357 struct sk_buff *skb)
1358{
1359 struct nlattr *start;
1360
1361 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1362 if (!start)
1363 return -EMSGSIZE;
1364
Joe Stringerab38a7b2015-10-06 11:00:01 -07001365 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
Joe Stringer7f8a4362015-08-26 11:31:48 -07001366 return -EMSGSIZE;
1367 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1368 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1369 return -EMSGSIZE;
Joe Stringere754ec62015-10-19 19:19:00 -07001370 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
Joe Stringer182e3042015-08-26 11:31:49 -07001371 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1372 &ct_info->mark))
1373 return -EMSGSIZE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001374 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringere754ec62015-10-19 19:19:00 -07001375 labels_nonzero(&ct_info->labels.mask) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001376 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1377 &ct_info->labels))
Joe Stringerc2ac6672015-08-26 11:31:52 -07001378 return -EMSGSIZE;
Joe Stringercae3a262015-08-26 11:31:53 -07001379 if (ct_info->helper) {
1380 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1381 ct_info->helper->name))
1382 return -EMSGSIZE;
1383 }
Jarno Rajahalme05752522016-03-10 10:54:23 -08001384#ifdef CONFIG_NF_NAT_NEEDED
1385 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1386 return -EMSGSIZE;
1387#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001388 nla_nest_end(skb, start);
1389
1390 return 0;
1391}
1392
1393void ovs_ct_free_action(const struct nlattr *a)
1394{
1395 struct ovs_conntrack_info *ct_info = nla_data(a);
1396
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001397 __ovs_ct_free_action(ct_info);
1398}
1399
1400static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1401{
Joe Stringercae3a262015-08-26 11:31:53 -07001402 if (ct_info->helper)
1403 module_put(ct_info->helper->me);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001404 if (ct_info->ct)
Joe Stringer76644232016-09-01 18:01:47 -07001405 nf_ct_tmpl_free(ct_info->ct);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001406}
Joe Stringerc2ac6672015-08-26 11:31:52 -07001407
1408void ovs_ct_init(struct net *net)
1409{
Joe Stringer33db4122015-10-01 15:00:37 -07001410 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001411 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1412
Florian Westphaladff6c62016-04-12 18:14:25 +02001413 if (nf_connlabels_get(net, n_bits - 1)) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001414 ovs_net->xt_label = false;
1415 OVS_NLERR(true, "Failed to set connlabel length");
1416 } else {
1417 ovs_net->xt_label = true;
1418 }
1419}
1420
1421void ovs_ct_exit(struct net *net)
1422{
1423 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1424
1425 if (ovs_net->xt_label)
1426 nf_connlabels_put(net);
1427}