blob: eab210bb1ef0f5ad6b93f8a2e597f47d789b57b9 [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
Joe Stringer49e261a2016-04-18 14:51:47 -0700370 skb_orphan(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700371 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100372 err = nf_ct_frag6_gather(net, skb, user);
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800373 if (err) {
374 if (err != -EINPROGRESS)
375 kfree_skb(skb);
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100376 return err;
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800377 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700378
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100379 key->ip.proto = ipv6_hdr(skb)->nexthdr;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700380 ovs_cb.mru = IP6CB(skb)->frag_max_size;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700381#endif
382 } else {
Joe Stringer74c16612015-10-25 20:21:48 -0700383 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700384 return -EPFNOSUPPORT;
385 }
386
387 key->ip.frag = OVS_FRAG_TYPE_NONE;
388 skb_clear_hash(skb);
389 skb->ignore_df = 1;
390 *OVS_CB(skb) = ovs_cb;
391
392 return 0;
393}
394
395static struct nf_conntrack_expect *
396ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
397 u16 proto, const struct sk_buff *skb)
398{
399 struct nf_conntrack_tuple tuple;
400
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;
403 return __nf_ct_expect_find(net, zone, &tuple);
404}
405
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800406/* This replicates logic from nf_conntrack_core.c that is not exported. */
407static enum ip_conntrack_info
408ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
409{
410 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
411
412 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
413 return IP_CT_ESTABLISHED_REPLY;
414 /* Once we've had two way comms, always ESTABLISHED. */
415 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
416 return IP_CT_ESTABLISHED;
417 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
418 return IP_CT_RELATED;
419 return IP_CT_NEW;
420}
421
422/* Find an existing connection which this packet belongs to without
423 * re-attributing statistics or modifying the connection state. This allows an
424 * skb->nfct lost due to an upcall to be recovered during actions execution.
425 *
426 * Must be called with rcu_read_lock.
427 *
428 * On success, populates skb->nfct and skb->nfctinfo, and returns the
429 * connection. Returns NULL if there is no existing entry.
430 */
431static struct nf_conn *
432ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
433 u8 l3num, struct sk_buff *skb)
434{
435 struct nf_conntrack_l3proto *l3proto;
436 struct nf_conntrack_l4proto *l4proto;
437 struct nf_conntrack_tuple tuple;
438 struct nf_conntrack_tuple_hash *h;
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800439 struct nf_conn *ct;
440 unsigned int dataoff;
441 u8 protonum;
442
443 l3proto = __nf_ct_l3proto_find(l3num);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800444 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
445 &protonum) <= 0) {
446 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
447 return NULL;
448 }
449 l4proto = __nf_ct_l4proto_find(l3num, protonum);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800450 if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
451 protonum, net, &tuple, l3proto, l4proto)) {
452 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
453 return NULL;
454 }
455
456 /* look for tuple match */
457 h = nf_conntrack_find_get(net, zone, &tuple);
458 if (!h)
459 return NULL; /* Not found. */
460
461 ct = nf_ct_tuplehash_to_ctrack(h);
462
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800463 skb->nfct = &ct->ct_general;
Jarno Rajahalmebce91f82016-08-01 19:36:07 -0700464 skb->nfctinfo = ovs_ct_get_info(h);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800465 return ct;
466}
467
Joe Stringer7f8a4362015-08-26 11:31:48 -0700468/* Determine whether skb->nfct is equal to the result of conntrack lookup. */
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800469static bool skb_nfct_cached(struct net *net,
470 const struct sw_flow_key *key,
471 const struct ovs_conntrack_info *info,
472 struct sk_buff *skb)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700473{
474 enum ip_conntrack_info ctinfo;
475 struct nf_conn *ct;
476
477 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800478 /* If no ct, check if we have evidence that an existing conntrack entry
479 * might be found for this skb. This happens when we lose a skb->nfct
480 * due to an upcall. If the connection was not confirmed, it is not
481 * cached and needs to be run through conntrack again.
482 */
483 if (!ct && key->ct.state & OVS_CS_F_TRACKED &&
484 !(key->ct.state & OVS_CS_F_INVALID) &&
485 key->ct.zone == info->zone.id)
486 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700487 if (!ct)
488 return false;
489 if (!net_eq(net, read_pnet(&ct->ct_net)))
490 return false;
491 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
492 return false;
Joe Stringercae3a262015-08-26 11:31:53 -0700493 if (info->helper) {
494 struct nf_conn_help *help;
495
496 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
497 if (help && rcu_access_pointer(help->helper) != info->helper)
498 return false;
499 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700500
501 return true;
502}
503
Jarno Rajahalme05752522016-03-10 10:54:23 -0800504#ifdef CONFIG_NF_NAT_NEEDED
505/* Modelled after nf_nat_ipv[46]_fn().
506 * range is only used for new, uninitialized NAT state.
507 * Returns either NF_ACCEPT or NF_DROP.
508 */
509static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
510 enum ip_conntrack_info ctinfo,
511 const struct nf_nat_range *range,
512 enum nf_nat_manip_type maniptype)
513{
514 int hooknum, nh_off, err = NF_ACCEPT;
515
516 nh_off = skb_network_offset(skb);
Lance Richardson18767ac2017-01-12 19:33:18 -0500517 skb_pull_rcsum(skb, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800518
519 /* See HOOK2MANIP(). */
520 if (maniptype == NF_NAT_MANIP_SRC)
521 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
522 else
523 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
524
525 switch (ctinfo) {
526 case IP_CT_RELATED:
527 case IP_CT_RELATED_REPLY:
Arnd Bergmann99b72482016-03-18 14:33:45 +0100528 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
529 skb->protocol == htons(ETH_P_IP) &&
Jarno Rajahalme05752522016-03-10 10:54:23 -0800530 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
531 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
532 hooknum))
533 err = NF_DROP;
534 goto push;
Arnd Bergmann99b72482016-03-18 14:33:45 +0100535 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
536 skb->protocol == htons(ETH_P_IPV6)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800537 __be16 frag_off;
538 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
539 int hdrlen = ipv6_skip_exthdr(skb,
540 sizeof(struct ipv6hdr),
541 &nexthdr, &frag_off);
542
543 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
544 if (!nf_nat_icmpv6_reply_translation(skb, ct,
545 ctinfo,
546 hooknum,
547 hdrlen))
548 err = NF_DROP;
549 goto push;
550 }
Jarno Rajahalme05752522016-03-10 10:54:23 -0800551 }
552 /* Non-ICMP, fall thru to initialize if needed. */
553 case IP_CT_NEW:
554 /* Seen it before? This can happen for loopback, retrans,
555 * or local packets.
556 */
557 if (!nf_nat_initialized(ct, maniptype)) {
558 /* Initialize according to the NAT action. */
559 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
560 /* Action is set up to establish a new
561 * mapping.
562 */
563 ? nf_nat_setup_info(ct, range, maniptype)
564 : nf_nat_alloc_null_binding(ct, hooknum);
565 if (err != NF_ACCEPT)
566 goto push;
567 }
568 break;
569
570 case IP_CT_ESTABLISHED:
571 case IP_CT_ESTABLISHED_REPLY:
572 break;
573
574 default:
575 err = NF_DROP;
576 goto push;
577 }
578
579 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
580push:
581 skb_push(skb, nh_off);
Lance Richardson18767ac2017-01-12 19:33:18 -0500582 skb_postpush_rcsum(skb, skb->data, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800583
584 return err;
585}
586
587static void ovs_nat_update_key(struct sw_flow_key *key,
588 const struct sk_buff *skb,
589 enum nf_nat_manip_type maniptype)
590{
591 if (maniptype == NF_NAT_MANIP_SRC) {
592 __be16 src;
593
594 key->ct.state |= OVS_CS_F_SRC_NAT;
595 if (key->eth.type == htons(ETH_P_IP))
596 key->ipv4.addr.src = ip_hdr(skb)->saddr;
597 else if (key->eth.type == htons(ETH_P_IPV6))
598 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
599 sizeof(key->ipv6.addr.src));
600 else
601 return;
602
603 if (key->ip.proto == IPPROTO_UDP)
604 src = udp_hdr(skb)->source;
605 else if (key->ip.proto == IPPROTO_TCP)
606 src = tcp_hdr(skb)->source;
607 else if (key->ip.proto == IPPROTO_SCTP)
608 src = sctp_hdr(skb)->source;
609 else
610 return;
611
612 key->tp.src = src;
613 } else {
614 __be16 dst;
615
616 key->ct.state |= OVS_CS_F_DST_NAT;
617 if (key->eth.type == htons(ETH_P_IP))
618 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
619 else if (key->eth.type == htons(ETH_P_IPV6))
620 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
621 sizeof(key->ipv6.addr.dst));
622 else
623 return;
624
625 if (key->ip.proto == IPPROTO_UDP)
626 dst = udp_hdr(skb)->dest;
627 else if (key->ip.proto == IPPROTO_TCP)
628 dst = tcp_hdr(skb)->dest;
629 else if (key->ip.proto == IPPROTO_SCTP)
630 dst = sctp_hdr(skb)->dest;
631 else
632 return;
633
634 key->tp.dst = dst;
635 }
636}
637
638/* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
639static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
640 const struct ovs_conntrack_info *info,
641 struct sk_buff *skb, struct nf_conn *ct,
642 enum ip_conntrack_info ctinfo)
643{
644 enum nf_nat_manip_type maniptype;
645 int err;
646
647 if (nf_ct_is_untracked(ct)) {
648 /* A NAT action may only be performed on tracked packets. */
649 return NF_ACCEPT;
650 }
651
652 /* Add NAT extension if not confirmed yet. */
653 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
654 return NF_ACCEPT; /* Can't NAT. */
655
656 /* Determine NAT type.
657 * Check if the NAT type can be deduced from the tracked connection.
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700658 * Make sure new expected connections (IP_CT_RELATED) are NATted only
659 * when committing.
Jarno Rajahalme05752522016-03-10 10:54:23 -0800660 */
661 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
662 ct->status & IPS_NAT_MASK &&
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700663 (ctinfo != IP_CT_RELATED || info->commit)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800664 /* NAT an established or related connection like before. */
665 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
666 /* This is the REPLY direction for a connection
667 * for which NAT was applied in the forward
668 * direction. Do the reverse NAT.
669 */
670 maniptype = ct->status & IPS_SRC_NAT
671 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
672 else
673 maniptype = ct->status & IPS_SRC_NAT
674 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
675 } else if (info->nat & OVS_CT_SRC_NAT) {
676 maniptype = NF_NAT_MANIP_SRC;
677 } else if (info->nat & OVS_CT_DST_NAT) {
678 maniptype = NF_NAT_MANIP_DST;
679 } else {
680 return NF_ACCEPT; /* Connection is not NATed. */
681 }
682 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
683
684 /* Mark NAT done if successful and update the flow key. */
685 if (err == NF_ACCEPT)
686 ovs_nat_update_key(key, skb, maniptype);
687
688 return err;
689}
690#else /* !CONFIG_NF_NAT_NEEDED */
691static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
692 const struct ovs_conntrack_info *info,
693 struct sk_buff *skb, struct nf_conn *ct,
694 enum ip_conntrack_info ctinfo)
695{
696 return NF_ACCEPT;
697}
698#endif
699
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800700/* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800701 * not done already. Update key with new CT state after passing the packet
702 * through conntrack.
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800703 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
704 * set to NULL and 0 will be returned.
705 */
Joe Stringer4f0909e2015-10-19 19:18:59 -0700706static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
Joe Stringer7f8a4362015-08-26 11:31:48 -0700707 const struct ovs_conntrack_info *info,
708 struct sk_buff *skb)
709{
710 /* If we are recirculating packets to match on conntrack fields and
711 * committing with a separate conntrack action, then we don't need to
712 * actually run the packet through conntrack twice unless it's for a
713 * different zone.
714 */
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800715 bool cached = skb_nfct_cached(net, key, info, skb);
716 enum ip_conntrack_info ctinfo;
717 struct nf_conn *ct;
718
719 if (!cached) {
Joe Stringer7f8a4362015-08-26 11:31:48 -0700720 struct nf_conn *tmpl = info->ct;
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800721 int err;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700722
723 /* Associate skb with specified zone. */
724 if (tmpl) {
725 if (skb->nfct)
726 nf_conntrack_put(skb->nfct);
727 nf_conntrack_get(&tmpl->ct_general);
728 skb->nfct = &tmpl->ct_general;
729 skb->nfctinfo = IP_CT_NEW;
730 }
731
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800732 /* Repeat if requested, see nf_iterate(). */
733 do {
734 err = nf_conntrack_in(net, info->family,
735 NF_INET_PRE_ROUTING, skb);
736 } while (err == NF_REPEAT);
737
738 if (err != NF_ACCEPT)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700739 return -ENOENT;
Joe Stringercae3a262015-08-26 11:31:53 -0700740
Jarno Rajahalme05752522016-03-10 10:54:23 -0800741 /* Clear CT state NAT flags to mark that we have not yet done
742 * NAT after the nf_conntrack_in() call. We can actually clear
743 * the whole state, as it will be re-initialized below.
744 */
745 key->ct.state = 0;
746
747 /* Update the key, but keep the NAT flags. */
748 ovs_ct_update_key(skb, info, key, true, true);
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800749 }
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800750
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800751 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800752 if (ct) {
753 /* Packets starting a new connection must be NATted before the
754 * helper, so that the helper knows about the NAT. We enforce
755 * this by delaying both NAT and helper calls for unconfirmed
756 * connections until the committing CT action. For later
757 * packets NAT and Helper may be called in either order.
758 *
759 * NAT will be done only if the CT action has NAT, and only
760 * once per packet (per zone), as guarded by the NAT bits in
761 * the key->ct.state.
762 */
763 if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) &&
764 (nf_ct_is_confirmed(ct) || info->commit) &&
765 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
766 return -EINVAL;
767 }
768
Joe Stringer16ec3d42016-05-11 10:29:26 -0700769 /* Userspace may decide to perform a ct lookup without a helper
770 * specified followed by a (recirculate and) commit with one.
771 * Therefore, for unconfirmed connections which we will commit,
772 * we need to attach the helper here.
773 */
774 if (!nf_ct_is_confirmed(ct) && info->commit &&
775 info->helper && !nfct_help(ct)) {
776 int err = __nf_ct_try_assign_helper(ct, info->ct,
777 GFP_ATOMIC);
778 if (err)
779 return err;
780 }
781
Jarno Rajahalme05752522016-03-10 10:54:23 -0800782 /* Call the helper only if:
783 * - nf_conntrack_in() was executed above ("!cached") for a
784 * confirmed connection, or
785 * - When committing an unconfirmed connection.
786 */
787 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
788 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
789 return -EINVAL;
790 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700791 }
792
793 return 0;
794}
795
796/* Lookup connection and read fields into key. */
797static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
798 const struct ovs_conntrack_info *info,
799 struct sk_buff *skb)
800{
801 struct nf_conntrack_expect *exp;
802
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800803 /* If we pass an expected packet through nf_conntrack_in() the
804 * expectation is typically removed, but the packet could still be
805 * lost in upcall processing. To prevent this from happening we
806 * perform an explicit expectation lookup. Expected connections are
807 * always new, and will be passed through conntrack only when they are
808 * committed, as it is OK to remove the expectation at that time.
809 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700810 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
811 if (exp) {
812 u8 state;
813
Jarno Rajahalme05752522016-03-10 10:54:23 -0800814 /* NOTE: New connections are NATted and Helped only when
815 * committed, so we are not calling into NAT here.
816 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700817 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
Joe Stringer182e3042015-08-26 11:31:49 -0700818 __ovs_ct_update_key(key, state, &info->zone, exp->master);
Samuel Gauthierd913d3a2016-06-28 17:22:26 +0200819 } else {
820 struct nf_conn *ct;
821 int err;
822
823 err = __ovs_ct_lookup(net, key, info, skb);
824 if (err)
825 return err;
826
827 ct = (struct nf_conn *)skb->nfct;
828 if (ct)
829 nf_ct_deliver_cached_events(ct);
830 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700831
832 return 0;
833}
834
Joe Stringer33db4122015-10-01 15:00:37 -0700835static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700836{
837 size_t i;
838
Joe Stringer33db4122015-10-01 15:00:37 -0700839 for (i = 0; i < sizeof(*labels); i++)
840 if (labels->ct_labels[i])
Joe Stringerc2ac6672015-08-26 11:31:52 -0700841 return true;
842
843 return false;
844}
845
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700846/* Lookup connection and confirm if unconfirmed. */
847static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
848 const struct ovs_conntrack_info *info,
849 struct sk_buff *skb)
850{
851 int err;
852
853 err = __ovs_ct_lookup(net, key, info, skb);
854 if (err)
855 return err;
856
857 /* Apply changes before confirming the connection so that the initial
858 * conntrack NEW netlink event carries the values given in the CT
859 * action.
860 */
861 if (info->mark.mask) {
862 err = ovs_ct_set_mark(skb, key, info->mark.value,
863 info->mark.mask);
864 if (err)
865 return err;
866 }
867 if (labels_nonzero(&info->labels.mask)) {
868 err = ovs_ct_set_labels(skb, key, &info->labels.value,
869 &info->labels.mask);
870 if (err)
871 return err;
872 }
873 /* This will take care of sending queued events even if the connection
874 * is already confirmed.
875 */
876 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
877 return -EINVAL;
878
879 return 0;
880}
881
Joe Stringer74c16612015-10-25 20:21:48 -0700882/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
883 * value if 'skb' is freed.
884 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700885int ovs_ct_execute(struct net *net, struct sk_buff *skb,
886 struct sw_flow_key *key,
887 const struct ovs_conntrack_info *info)
888{
889 int nh_ofs;
890 int err;
891
892 /* The conntrack module expects to be working at L3. */
893 nh_ofs = skb_network_offset(skb);
Lance Richardson18767ac2017-01-12 19:33:18 -0500894 skb_pull_rcsum(skb, nh_ofs);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700895
896 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
897 err = handle_fragments(net, key, info->zone.id, skb);
898 if (err)
899 return err;
900 }
901
Joe Stringerab38a7b2015-10-06 11:00:01 -0700902 if (info->commit)
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700903 err = ovs_ct_commit(net, key, info, skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700904 else
905 err = ovs_ct_lookup(net, key, info, skb);
906
907 skb_push(skb, nh_ofs);
Lance Richardson18767ac2017-01-12 19:33:18 -0500908 skb_postpush_rcsum(skb, skb->data, nh_ofs);
Joe Stringer74c16612015-10-25 20:21:48 -0700909 if (err)
910 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700911 return err;
912}
913
Joe Stringercae3a262015-08-26 11:31:53 -0700914static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
915 const struct sw_flow_key *key, bool log)
916{
917 struct nf_conntrack_helper *helper;
918 struct nf_conn_help *help;
919
920 helper = nf_conntrack_helper_try_module_get(name, info->family,
921 key->ip.proto);
922 if (!helper) {
923 OVS_NLERR(log, "Unknown helper \"%s\"", name);
924 return -EINVAL;
925 }
926
927 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
928 if (!help) {
929 module_put(helper->me);
930 return -ENOMEM;
931 }
932
933 rcu_assign_pointer(help->helper, helper);
934 info->helper = helper;
935 return 0;
936}
937
Jarno Rajahalme05752522016-03-10 10:54:23 -0800938#ifdef CONFIG_NF_NAT_NEEDED
939static int parse_nat(const struct nlattr *attr,
940 struct ovs_conntrack_info *info, bool log)
941{
942 struct nlattr *a;
943 int rem;
944 bool have_ip_max = false;
945 bool have_proto_max = false;
946 bool ip_vers = (info->family == NFPROTO_IPV6);
947
948 nla_for_each_nested(a, attr, rem) {
949 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
950 [OVS_NAT_ATTR_SRC] = {0, 0},
951 [OVS_NAT_ATTR_DST] = {0, 0},
952 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
953 sizeof(struct in6_addr)},
954 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
955 sizeof(struct in6_addr)},
956 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
957 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
958 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
959 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
960 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
961 };
962 int type = nla_type(a);
963
964 if (type > OVS_NAT_ATTR_MAX) {
965 OVS_NLERR(log,
966 "Unknown NAT attribute (type=%d, max=%d).\n",
967 type, OVS_NAT_ATTR_MAX);
968 return -EINVAL;
969 }
970
971 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
972 OVS_NLERR(log,
973 "NAT attribute type %d has unexpected length (%d != %d).\n",
974 type, nla_len(a),
975 ovs_nat_attr_lens[type][ip_vers]);
976 return -EINVAL;
977 }
978
979 switch (type) {
980 case OVS_NAT_ATTR_SRC:
981 case OVS_NAT_ATTR_DST:
982 if (info->nat) {
983 OVS_NLERR(log,
984 "Only one type of NAT may be specified.\n"
985 );
986 return -ERANGE;
987 }
988 info->nat |= OVS_CT_NAT;
989 info->nat |= ((type == OVS_NAT_ATTR_SRC)
990 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
991 break;
992
993 case OVS_NAT_ATTR_IP_MIN:
Haishuang Yanac71b462016-03-28 18:08:59 +0800994 nla_memcpy(&info->range.min_addr, a,
995 sizeof(info->range.min_addr));
Jarno Rajahalme05752522016-03-10 10:54:23 -0800996 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
997 break;
998
999 case OVS_NAT_ATTR_IP_MAX:
1000 have_ip_max = true;
1001 nla_memcpy(&info->range.max_addr, a,
1002 sizeof(info->range.max_addr));
1003 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1004 break;
1005
1006 case OVS_NAT_ATTR_PROTO_MIN:
1007 info->range.min_proto.all = htons(nla_get_u16(a));
1008 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1009 break;
1010
1011 case OVS_NAT_ATTR_PROTO_MAX:
1012 have_proto_max = true;
1013 info->range.max_proto.all = htons(nla_get_u16(a));
1014 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1015 break;
1016
1017 case OVS_NAT_ATTR_PERSISTENT:
1018 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1019 break;
1020
1021 case OVS_NAT_ATTR_PROTO_HASH:
1022 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1023 break;
1024
1025 case OVS_NAT_ATTR_PROTO_RANDOM:
1026 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1027 break;
1028
1029 default:
1030 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1031 return -EINVAL;
1032 }
1033 }
1034
1035 if (rem > 0) {
1036 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1037 return -EINVAL;
1038 }
1039 if (!info->nat) {
1040 /* Do not allow flags if no type is given. */
1041 if (info->range.flags) {
1042 OVS_NLERR(log,
1043 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1044 );
1045 return -EINVAL;
1046 }
1047 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1048 } else if (!info->commit) {
1049 OVS_NLERR(log,
1050 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1051 );
1052 return -EINVAL;
1053 }
1054 /* Allow missing IP_MAX. */
1055 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1056 memcpy(&info->range.max_addr, &info->range.min_addr,
1057 sizeof(info->range.max_addr));
1058 }
1059 /* Allow missing PROTO_MAX. */
1060 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1061 !have_proto_max) {
1062 info->range.max_proto.all = info->range.min_proto.all;
1063 }
1064 return 0;
1065}
1066#endif
1067
Joe Stringer7f8a4362015-08-26 11:31:48 -07001068static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001069 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
Joe Stringer7f8a4362015-08-26 11:31:48 -07001070 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1071 .maxlen = sizeof(u16) },
Joe Stringer182e3042015-08-26 11:31:49 -07001072 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1073 .maxlen = sizeof(struct md_mark) },
Joe Stringer33db4122015-10-01 15:00:37 -07001074 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1075 .maxlen = sizeof(struct md_labels) },
Joe Stringercae3a262015-08-26 11:31:53 -07001076 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
Jarno Rajahalme05752522016-03-10 10:54:23 -08001077 .maxlen = NF_CT_HELPER_NAME_LEN },
1078#ifdef CONFIG_NF_NAT_NEEDED
1079 /* NAT length is checked when parsing the nested attributes. */
1080 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1081#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001082};
1083
1084static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
Joe Stringercae3a262015-08-26 11:31:53 -07001085 const char **helper, bool log)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001086{
1087 struct nlattr *a;
1088 int rem;
1089
1090 nla_for_each_nested(a, attr, rem) {
1091 int type = nla_type(a);
1092 int maxlen = ovs_ct_attr_lens[type].maxlen;
1093 int minlen = ovs_ct_attr_lens[type].minlen;
1094
1095 if (type > OVS_CT_ATTR_MAX) {
1096 OVS_NLERR(log,
1097 "Unknown conntrack attr (type=%d, max=%d)",
1098 type, OVS_CT_ATTR_MAX);
1099 return -EINVAL;
1100 }
1101 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1102 OVS_NLERR(log,
1103 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1104 type, nla_len(a), maxlen);
1105 return -EINVAL;
1106 }
1107
1108 switch (type) {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001109 case OVS_CT_ATTR_COMMIT:
1110 info->commit = true;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001111 break;
1112#ifdef CONFIG_NF_CONNTRACK_ZONES
1113 case OVS_CT_ATTR_ZONE:
1114 info->zone.id = nla_get_u16(a);
1115 break;
1116#endif
Joe Stringer182e3042015-08-26 11:31:49 -07001117#ifdef CONFIG_NF_CONNTRACK_MARK
1118 case OVS_CT_ATTR_MARK: {
1119 struct md_mark *mark = nla_data(a);
1120
Joe Stringere754ec62015-10-19 19:19:00 -07001121 if (!mark->mask) {
1122 OVS_NLERR(log, "ct_mark mask cannot be 0");
1123 return -EINVAL;
1124 }
Joe Stringer182e3042015-08-26 11:31:49 -07001125 info->mark = *mark;
1126 break;
1127 }
1128#endif
Joe Stringerc2ac6672015-08-26 11:31:52 -07001129#ifdef CONFIG_NF_CONNTRACK_LABELS
Joe Stringer33db4122015-10-01 15:00:37 -07001130 case OVS_CT_ATTR_LABELS: {
1131 struct md_labels *labels = nla_data(a);
Joe Stringerc2ac6672015-08-26 11:31:52 -07001132
Joe Stringere754ec62015-10-19 19:19:00 -07001133 if (!labels_nonzero(&labels->mask)) {
1134 OVS_NLERR(log, "ct_labels mask cannot be 0");
1135 return -EINVAL;
1136 }
Joe Stringer33db4122015-10-01 15:00:37 -07001137 info->labels = *labels;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001138 break;
1139 }
1140#endif
Joe Stringercae3a262015-08-26 11:31:53 -07001141 case OVS_CT_ATTR_HELPER:
1142 *helper = nla_data(a);
1143 if (!memchr(*helper, '\0', nla_len(a))) {
1144 OVS_NLERR(log, "Invalid conntrack helper");
1145 return -EINVAL;
1146 }
1147 break;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001148#ifdef CONFIG_NF_NAT_NEEDED
1149 case OVS_CT_ATTR_NAT: {
1150 int err = parse_nat(a, info, log);
1151
1152 if (err)
1153 return err;
1154 break;
1155 }
1156#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001157 default:
1158 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1159 type);
1160 return -EINVAL;
1161 }
1162 }
1163
Jarno Rajahalme7d904c72016-06-21 14:59:38 -07001164#ifdef CONFIG_NF_CONNTRACK_MARK
1165 if (!info->commit && info->mark.mask) {
1166 OVS_NLERR(log,
1167 "Setting conntrack mark requires 'commit' flag.");
1168 return -EINVAL;
1169 }
1170#endif
1171#ifdef CONFIG_NF_CONNTRACK_LABELS
1172 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1173 OVS_NLERR(log,
1174 "Setting conntrack labels requires 'commit' flag.");
1175 return -EINVAL;
1176 }
1177#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001178 if (rem > 0) {
1179 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1180 return -EINVAL;
1181 }
1182
1183 return 0;
1184}
1185
Joe Stringerc2ac6672015-08-26 11:31:52 -07001186bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001187{
1188 if (attr == OVS_KEY_ATTR_CT_STATE)
1189 return true;
1190 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1191 attr == OVS_KEY_ATTR_CT_ZONE)
1192 return true;
Joe Stringer182e3042015-08-26 11:31:49 -07001193 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1194 attr == OVS_KEY_ATTR_CT_MARK)
1195 return true;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001196 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001197 attr == OVS_KEY_ATTR_CT_LABELS) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001198 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1199
1200 return ovs_net->xt_label;
1201 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001202
1203 return false;
1204}
1205
1206int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1207 const struct sw_flow_key *key,
1208 struct sw_flow_actions **sfa, bool log)
1209{
1210 struct ovs_conntrack_info ct_info;
Joe Stringercae3a262015-08-26 11:31:53 -07001211 const char *helper = NULL;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001212 u16 family;
1213 int err;
1214
1215 family = key_to_nfproto(key);
1216 if (family == NFPROTO_UNSPEC) {
1217 OVS_NLERR(log, "ct family unspecified");
1218 return -EINVAL;
1219 }
1220
1221 memset(&ct_info, 0, sizeof(ct_info));
1222 ct_info.family = family;
1223
1224 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1225 NF_CT_DEFAULT_ZONE_DIR, 0);
1226
Joe Stringercae3a262015-08-26 11:31:53 -07001227 err = parse_ct(attr, &ct_info, &helper, log);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001228 if (err)
1229 return err;
1230
1231 /* Set up template for tracking connections in specific zones. */
1232 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1233 if (!ct_info.ct) {
1234 OVS_NLERR(log, "Failed to allocate conntrack template");
1235 return -ENOMEM;
1236 }
Joe Stringer90c7afc962015-12-23 14:39:27 -08001237
1238 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1239 nf_conntrack_get(&ct_info.ct->ct_general);
1240
Joe Stringercae3a262015-08-26 11:31:53 -07001241 if (helper) {
1242 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1243 if (err)
1244 goto err_free_ct;
1245 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001246
1247 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1248 sizeof(ct_info), log);
1249 if (err)
1250 goto err_free_ct;
1251
Joe Stringer7f8a4362015-08-26 11:31:48 -07001252 return 0;
1253err_free_ct:
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001254 __ovs_ct_free_action(&ct_info);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001255 return err;
1256}
1257
Jarno Rajahalme05752522016-03-10 10:54:23 -08001258#ifdef CONFIG_NF_NAT_NEEDED
1259static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1260 struct sk_buff *skb)
1261{
1262 struct nlattr *start;
1263
1264 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1265 if (!start)
1266 return false;
1267
1268 if (info->nat & OVS_CT_SRC_NAT) {
1269 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1270 return false;
1271 } else if (info->nat & OVS_CT_DST_NAT) {
1272 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1273 return false;
1274 } else {
1275 goto out;
1276 }
1277
1278 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
Arnd Bergmann99b72482016-03-18 14:33:45 +01001279 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1280 info->family == NFPROTO_IPV4) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001281 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1282 info->range.min_addr.ip) ||
1283 (info->range.max_addr.ip
1284 != info->range.min_addr.ip &&
1285 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1286 info->range.max_addr.ip))))
1287 return false;
Arnd Bergmann99b72482016-03-18 14:33:45 +01001288 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1289 info->family == NFPROTO_IPV6) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001290 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1291 &info->range.min_addr.in6) ||
1292 (memcmp(&info->range.max_addr.in6,
1293 &info->range.min_addr.in6,
1294 sizeof(info->range.max_addr.in6)) &&
1295 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1296 &info->range.max_addr.in6))))
1297 return false;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001298 } else {
1299 return false;
1300 }
1301 }
1302 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1303 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1304 ntohs(info->range.min_proto.all)) ||
1305 (info->range.max_proto.all != info->range.min_proto.all &&
1306 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1307 ntohs(info->range.max_proto.all)))))
1308 return false;
1309
1310 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1311 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1312 return false;
1313 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1314 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1315 return false;
1316 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1317 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1318 return false;
1319out:
1320 nla_nest_end(skb, start);
1321
1322 return true;
1323}
1324#endif
1325
Joe Stringer7f8a4362015-08-26 11:31:48 -07001326int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1327 struct sk_buff *skb)
1328{
1329 struct nlattr *start;
1330
1331 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1332 if (!start)
1333 return -EMSGSIZE;
1334
Joe Stringerab38a7b2015-10-06 11:00:01 -07001335 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
Joe Stringer7f8a4362015-08-26 11:31:48 -07001336 return -EMSGSIZE;
1337 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1338 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1339 return -EMSGSIZE;
Joe Stringere754ec62015-10-19 19:19:00 -07001340 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
Joe Stringer182e3042015-08-26 11:31:49 -07001341 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1342 &ct_info->mark))
1343 return -EMSGSIZE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001344 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringere754ec62015-10-19 19:19:00 -07001345 labels_nonzero(&ct_info->labels.mask) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001346 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1347 &ct_info->labels))
Joe Stringerc2ac6672015-08-26 11:31:52 -07001348 return -EMSGSIZE;
Joe Stringercae3a262015-08-26 11:31:53 -07001349 if (ct_info->helper) {
1350 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1351 ct_info->helper->name))
1352 return -EMSGSIZE;
1353 }
Jarno Rajahalme05752522016-03-10 10:54:23 -08001354#ifdef CONFIG_NF_NAT_NEEDED
1355 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1356 return -EMSGSIZE;
1357#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001358 nla_nest_end(skb, start);
1359
1360 return 0;
1361}
1362
1363void ovs_ct_free_action(const struct nlattr *a)
1364{
1365 struct ovs_conntrack_info *ct_info = nla_data(a);
1366
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001367 __ovs_ct_free_action(ct_info);
1368}
1369
1370static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1371{
Joe Stringercae3a262015-08-26 11:31:53 -07001372 if (ct_info->helper)
1373 module_put(ct_info->helper->me);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001374 if (ct_info->ct)
Joe Stringer76644232016-09-01 18:01:47 -07001375 nf_ct_tmpl_free(ct_info->ct);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001376}
Joe Stringerc2ac6672015-08-26 11:31:52 -07001377
1378void ovs_ct_init(struct net *net)
1379{
Joe Stringer33db4122015-10-01 15:00:37 -07001380 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001381 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1382
Florian Westphaladff6c62016-04-12 18:14:25 +02001383 if (nf_connlabels_get(net, n_bits - 1)) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001384 ovs_net->xt_label = false;
1385 OVS_NLERR(true, "Failed to set connlabel length");
1386 } else {
1387 ovs_net->xt_label = true;
1388 }
1389}
1390
1391void ovs_ct_exit(struct net *net)
1392{
1393 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1394
1395 if (ovs_net->xt_label)
1396 nf_connlabels_put(net);
1397}